Home

EViews Illustrated for Version 7

1. 25 Deconstructing the Workfile 26 Time to Type 27 Identity Noncrisis 32 Dated Series 34 The Import Business 37 Adding Data To An Existing Workfile Or Being Rectangular Doesn t Mean Being Inflexible 51 Among the Missing 56 Quick Review 57 Appendix Having A Good Time With Your Date 58 CHAPTER 3 GETTING THE MOST FROM LEAST SQUARES 61 A First Regression
2. 267 CHAPTER 10 PRELUDE TO PANEL AND POOL 269 Pooled or Paneled Population 269 Nuances 271 So What Are the Benefits of Using Pools and Panels 272 Quick P review 272 CHAPTER 11 PANEL WHAT S MY LINE 275 What s So Nifty About Panel Data 275 EViews Illustrated book Page iii Monday February 25 2013 10 06 AM iv Table of Contents Setting Up Panel Data 276 Panel Estimation 278 Pretty Panel Pictures
3. 341 2SLS 345 Generalized Method of Moments 347 Limited Dependent Variables 349 ARCH etc 351 Maximum Likelihood Rolling Your Own 355 System Estimation 357 Vector Autoregressions VAR 361 Quick Review 364 EViews Illustrated book Page iv Monday February 25 2013 10 06 AM Table of Contents v CHAPTER 15 SUPER MODELS 365 Your First Homework Bam Taken Up A Notch
4. 397 Objects and Commands 398 Workfile Backups 399 Updates A Small Thing 400 Updates A Big Thing 400 Ready To Take A Break 401 Help 401 Odd Ending 401 CHAPTER 18 OPTIONAL ENDING 403 Required Options 403 EViews Illustrated book Page v Monday February 25 2013 10 06 AM vi Table of Contents Option al Recommendations
5. 100 Numbers and Letters 100 Can We Have A Date 106 What Are Your Values 109 Relative Exotica 112 Quick Review 114 CHAPTER 5 PICTURE THIS 117 A Simple Soup To Nuts Graphing Example 117 A Graphic Description of the Creative Process 126 Picture One Series 131 Group Graphics 137 Let s Look At This From Another Angle
6. 223 Just Push the Forecast Button 223 Theory of Forecasting 225 Dynamic Versus Static Forecasting 228 Sample Forecast Samples 229 Facing the Unknown 231 Forecast Evaluation 232 Forecasting Beneath the Surface 235 Quick Review Forecasting 238 CHAPTER 9 PAGE AFTER PAGE AFTER PAGE 239 Pages Are Easy To Reach 239 Creating New Pages
7. 61 The Really Important Regression Results 64 The Pretty Important But Not So Important As the Last Section s Regression Results 66 A Multiple Regression Is Simple Too 73 Hypothesis Testing 74 Representing 78 What s Left After You ve Gotten the Most Out of Least Squares 78 EViews Illustrated book Page i Monday February 25 2013 10 06 AM ii Table of Contents Quick Review 81 CHAPTER 4 DATA THE TRANSFORMATIONAL EXPERIENCE 83 Your Basic Elementary Algebra 83 Simple Sample Says 95 Data Types Plain and Fancy
8. 155 To Summarize 156 Categorical Graphs 157 Togetherness of the Second Sort 164 Quick Review and Look Ahead 166 CHAPTER 6 INTIMACY WITH GRAPHIC OBJECTS 167 To Freeze Or Not To Freeze Redux 168 A Touch of Text 168 Shady Areas and No Worry Lines 171 Templates for Success 173 Point Me The Way 177 Your Data Another Sorta Way 178
9. 283 More Panel Estimation Techniques 284 One Dimensional Two Dimensional Panels 285 Fixed Effects With and Without the Social Contrivance of Panel Structure 287 Quick Review Panel 289 CHAPTER 12 EVERYONE INTO THE POOL 291 Getting Your Feet Wet 291 Playing in the Pool Data 297 Getting Out of the Pool 302 More Pool Estimation 304 Getting Data In and Out of the Pool 309 Quick Review Pools 312 CHAPTER 13 SERIAL CORR
10. 413 INDEX 415 EViews Illustrated book Page vi Monday February 25 2013 10 06 AM Foreword Sit back put up your feet and prepare for the E Views ticket of your life My goal in writing EViews Illustrated is that you the reader should have some fun You might have thought the goal would have been to teach you EViews Well it is but books about software can be awfully dry I don t learn much when I m bored and you probably don t either By keeping a light touch I hope to make this tour of EViews enjoyable as well as productive Most of the book is written as if you were seated in front of an EViews computer session and you and I were having a conversation Reading the book while running EViews is certainly recommended but it ll also work just fine to read pretending EViews is running while you re actually plunked down in your favorite arm chair EViews is a big program You don t need to learn all of it Do read Chapter 1 A Quick Walk Through to get started After that feel comfortable to pick and choose the parts you find most valuable EViews Illustrated for Version 8 is keyed to release 8 of EViews Most but not all of EViews Illustrated applies to release 4 as well EViews workfiles discussed in the book are available for download from www
11. 403 More Detailed Options 406 Window Behavior 407 Font Options 408 Frequency Conversion 408 Alpha Truncation 409 Spreadsheet Defaults 409 Workfile Storage Defaults 410 Estimation Defaults 411 File Locations 412 Graphics Defaults 412 Quick Review
12. 240 Renaming Deleting and Saving Pages 243 Multi Page Workfiles The Most Basic Motivation 244 Multiple Frequencies Multiple Pages 244 Links The Live Connection 251 Unlinking 254 Have A Match 255 Matching When The Identifiers Are Really Different 258 Contracted Data 262 Expanded Data 264 Having Contractions 266 Two Hints and A GotchYa 267 Quick Review
13. EViews Illustrated book Page 345 Monday February 25 2013 10 06 AM 346 Chapter 14 A Taste of Advanced Estimation we expect and Using monthly U S data in CPI_AND_UNEMPLOYMENT wf1 we could try to estimate this equa tion by least squares with the com mand ls inf c inf 1 unrate 1 Notice that the estimated coefficients are very different from the coeffi cients predicted by theory The coef ficient on future inflation is approximately 0 5 rather than 1 0 and the coefficient on unemployment is positive The econometric difficulty is that by using actual future inflation as a proxy for expected future inflation we introduce an errors in variables problem We ll try a 2SLS estimate using lagged information as instruments The 2SLS command uses the same equation specification as does least squares The equa tion specification is followed by an and the list of instruments The command name is tsls Econometric digression If a right hand side variable in a regression is measured with random error the equation is said to suffer from errors in variables Errors in variables leads to biased coefficients in ordinary least squares Sometimes this can be fixed with 2SLS Hint It s tsls rather than 2sls because the convention is that computer commands start with a letter rather than a number b 1 g 0 lt EViews Illustrated book Page 346 Monday February 25 2013 10 06 AM Gen
14. Hint First observation and Last observation are especially useful in the analysis of financial price data where point in time data are often preferred over time aggregated data EViews Illustrated book Page 250 Monday February 25 2013 10 06 AM Links The Live Connection 251 Paste Special Sometimes the default frequency conversion method isn t suitable Instead of using Paste to paste a series use Paste Special to bring up all the available conversion options Drag and dropping after right click selecting will also bring up the Paste Special menu You can choose the conversion method from the fields on the right In addition you can enter a new name for the pasted series in the field marked Pattern Nothing limits copy and paste or copy and paste special to a single series The Pattern field lets you specify a general pattern for changing the name of the pasted series For example the pattern _quarterly would paste the series X Y and Z as X_quarterly Y_quarterly and Z_quarterly For further discussion see the User s Guide Links The Live Connection Copies of data and the original source are related conceptually but mechanically they re completely unlinked Humans understand that a quarterly series for industrial production represents a view of an underlying monthly IP series But as a mechanical matter once the copying is done EViews no longer sees any connection between the quarterly data and the
15. EViews Illustrated book Page 331 Monday February 25 2013 10 06 AM 332 Chapter 13 Serial Correlation Friend or Foe That wasn t very hard The results are shown to the right Both ARMA coefficients are off the scale signifi cant The isn t very high but remember that we re explaining changes not levels of volume To estimate higher order ARIMA models just include more AR and MA terms in the command line ARIMA Forecasting Forecasting from an ARIMA model pretty much consists of pushing the button and then setting the options in the Forecast dialog as you would for any other equation You ll notice one new twist The dependent variable is D LOGVOL but EViews defaults to forecasting the level vari able LOGVOL R2 EViews Illustrated book Page 332 Monday February 25 2013 10 06 AM Quick Review 333 Here s our ARIMA based fore cast together with the regres sion based forecast generated earlier In the example at hand the ARIMA based forecast gets a bit closer to the true data Quick Review You can check for persistence in regression errors with a variety of visual aids as well as with formal statistical tests Regressions are easily corrected for the presence of ARMA errors by the addition of AR 1 AR 2 etc terms in the ls command ARMA and ARIMA models are easily estimated by using the ls command without including any exogenous variables on the right Forecasting requires nothing
16. looks like this Great for computers not so great for humans So EViews lets you change the display of a series containing date numbers In a spreadsheet view you can change the display by right clicking on a column and choosing Display format You can also open a series hit the button and change the Numeric display field to one of the date or time formats Then more fields will appear to let you further customize the format This looks a lot better EViews will also translate text strings into dates when doing an ASCII Read and set the initial display of the series read to be a date format So dates are pretty straightforward Except when they re not If you want more details the Command and Programming Reference has a very nice 20 page section for you Hint Since dates are stored as numbers you can do sensible date arithmetic If the series DATEBOUGHT and DATESOLD hold the information suggested by their respec tive names then series daysheld datesold datebought does just what it should EViews Illustrated book Page 59 Monday February 25 2013 10 06 AM 60 Chapter 2 EViews Meet Data EViews Illustrated book Page 60 Monday February 25 2013 10 06 AM Chapter 3 Getting the Most from Least Squares Regression is the king of econometric tools Regression s job is to find numerical values for theoretical parameters In the simplest case this means telling us the slope and intercept of a line drawn th
17. Pools have two special spreadsheet views stacked and unstacked chosen by pushing the button or choosing the View Spreadsheet stacked data menu For either view the first step is to specify the desired series when the Series List dialog opens Enter the names of the series you d like to see using the conventions that a series with a question mark means replace that question mark with each of the country ids in turn A series with no a question mark means use the series as usual repeated for each country The way we ve filled in the dialog here asks EViews to display D LOG POPCAN D LOG POPFRA etc for YCAN YFRA etc and for D LOG POPUSA separately Stacked View The spreadsheet opens with all the data for Canada followed by all the data for France etc The data for POPUSA gets repeated next to each country Notice how the identifier in the obs column gives the cross section identifier followed by the date in other words country and year This is called the stacked view You can imag ine putting together all the data for Canada then stacking on all the data for France etc We ll return to the idea of a stacked view when we talk about loading in pooled data below Hint It s fine to have multiple pool objects in the workfile They re just different lists of identifiers after all EViews Illustrated book Page 297 Monday February 25 2013 10 06 AM 298 Chapter 12 Everyone Into the Pool Unstac
18. The Edit axis menu controls whether the fields below apply to the left right bottom or top axis Switching the axis in the menu changes the fields in the dialog Left and right axes For axes scaled numerically the scaling method dropdown lets you pick between the standard linear scale a linear scale that s guaranteed to include zero a log scale and normalized data This last scale marks the mean of the data as zero and makes one vertical unit equivalent to one standard deviation Hint Most often it s the vertical axes that have numerical scaling dates being shown on the bottom But sometimes scatterplots are an example numerical scales appear on the x axis EViews Illustrated book Page 184 Monday February 25 2013 10 06 AM Options Options Options 185 The log scale is especially useful for data that exhibits roughly constant percentage growth As an example here s a plot of U S real GDP By plotting on a log scale we see a nice more or less straight line For the two vertical axes the Axis scale endpoints dropdown has choices for automatic data minimum and maximum and user specified Most of the time the automatic choice is fine but once in a while you may prefer to change the scale In addition on the Data axis labels page the Data units amp label format section allows you to label your axis using scaled units or if you wish to customize the formatting of your labels The Ticks dropdown provide
19. The following program uses a for loop to step through the forecast periods Notice how smpl is used first to control the estimation period and is then reset to the forecast period Temporary variables are used to get around the problem of the forecast procedure overwrit ing forecasts from earlier windows set window size window 20 get size of workfile length obsrange declare equation for estimation equation eq1 declare series for final results series yhat point estimates series yhat_se forecast std err set step size step 4 move sample step obs at a time for i 1 to length window 1 step step step set sample to estimation period smpl first i 1 first i window 2 estimate equation eq1 ls y c y 1 y 2 reset sample to forecast period smpl first i window 1 first i window 2 step make forecasts in temporary series first eq1 forecast f na tmp_yhat tmp_se copy data in current forecast sample yhat tmp_yhat yhat_se tmp_se next Monte Carlo Earlier in the chapter we briefly touched on the idea of a Monte Carlo study Here s a more in depth example A typical Monte Carlo simulation exercise consists of the following steps 1 Specify the true model data generating process underlying the data 2 Simulate a draw from the data and estimate the model using the simulated data 3 Repeat step 2 many times each time storing the results of interest EViews Ill
20. When appropriate dates are used in place of plain numbers EViews needs to know how observations are numbered When you set up a workfile the first thing you need to do is tell EViews how the identifier of your data is structured monthly annual just numbered 1 2 3 etc Your second task is to tell EViews the range your observations take January 1888 through January 2004 1939 through 1944 1 through 28 etc And that s all you need to know Hint When thinking of an econometric model a data series is often just called a vari able Hint Series columns don t have any inherent order but observation numbers rows do SALARY is neither before nor after DISCIPLINE in any important sense In contrast 2 really is the number after 1 Hint Every variable in an EViews workfile shares a common identifier series You can t have one variable that s measured in January February and March and a different variable that s measured in the chocolate mixing bowl the vanilla mixing bowl and the mocha mixing bowl Subhint Well yes actually you can EViews has quite sophisticated capabilities for handling both mixed frequency data and panel data These are covered later in the book EViews Illustrated book Page 24 Monday February 25 2013 10 06 AM Creating a New Workfile 25 Creating a New Workfile Open EViews and use the menu to choose File New Workfile The Workfile Create dialog pops up You ll
21. to bring up the dialog shown to the right In order to whip the Wald Test dialog into shape you need to know three things EViews names coefficients C 1 C 2 C 3 etc numbering them in the order they appear in the regression As an example the coeffi cient on LOG VOLUME 1 is C 4 You specify a hypothesis as an equation restricting the values of the coefficients in the regression To test that the coefficient on LOG VOLUME 1 equals zero specify C 4 0 If a hypothesis involves multiple restrictions you enter multiple coefficient equations separated by commas Let s work through some examples starting with the one we already know the answer to Is the coefficient on LOG VOLUME 1 significantly different from zero Hint We know the results of this test already because EViews computed the appropriate test statistic for us in its standard regression output EViews Illustrated book Page 75 Monday February 25 2013 10 06 AM 76 Chapter 3 Getting the Most from Least Squares Complete the Wald Test dialog with C 4 0 EViews gives the test results as shown to the right EViews always reports an F statistic since the F applies for both single and multiple restrictions In cases with a single restric tion EViews will also show the t statistic Suppose we wanted to test whether the coefficient on LOG VOLUME 1 equaled one rather than zero Enter c 4 1 to find the new test
22. 2 a EViews Illustrated book Page 64 Monday February 25 2013 10 06 AM The Really Important Regression Results 65 The convention for inline reporting works well for a single equation but becomes unwieldy when you have more than one equation to report Results from several related regressions might be displayed in a table looking something like Table 2 Column 2 Don t worry we ll come back to it later EViews regression output is divided into three panels The top panel summarizes the input to the regression the middle panel gives information about each regression coefficient and the bottom panel provides summary statistics about the whole regression equation Hint The dependent variable is also called the left hand side variable and the indepen dent variables are called the right hand side variables That s because when you write out the regression equation algebraically as above convention puts the dependent variable to the left of the equals sign and the independent variables to the right Table 2 1 2 Intercept 2 629649 0 089576 0 106396 0 045666 t 0 017278 0 000334 0 000736 0 000417 6 63E 06 1 37E 06 log volume 1 0 868273 0 022910 ser 0 967362 0 289391 0 852357 0 986826 Hint Good scientific practice is to report only digits that are meaningful when display ing a number We ve printed far too many digits in both the inline display and in
23. Click on the Mortality tab to activate the Mortal ity page Then create a new link named REV using the menu Object New Object Series Link The object appears in the workfile win dow with a pink background to indicate a link and a question mark showing that the link hasn t been specified Double clicking opens a view with an error message indicating that the properties of the link haven t been specified yet Click the button and then the Link Spec tab The default dia log asks about frequency conver sions which isn t what we need Click the General match merge criteria radio button in the Merge by field EViews Illustrated book Page 256 Monday February 25 2013 10 06 AM Have A Match 257 Notice that we now have a new set of fields on the right side of the dialog Fill out the dialog with the Source series and Workfile page in the Link to field Since we want to match observations that have the same state names enter STATE in both the Source index and Des tination index fields When you close the dialog you ll see that the link icon has switched to indicating that the link is now complete A quick glance at the data shows that EViews has made the correct obvious to us connec tion Hint You may find it more intuitive to think of the Link to field as the Link from field Remember that you re specifying the data source here The destination is always the active page EViews Illustrated book
24. In our example above we didn t have to face this issue because we were making an in sam ple forecast a forecast where We knew the values of all the explanatory variables for the forecast period because the forecast period was a subset of the estimation period In sample forecasting has two advantages you always have the required data and you can check the accuracy of your forecast by comparing it to what actually happened We forecast 11 79 percent currency growth for November 2004 Currency growth was actually 12 42 per cent so the forecast error was a bit above a half a percent Not too bad The alternative to in sample forecasting is out of sample forecasting where You have to obey the cardinal rule when forecasting out of sample And sometimes that s a problem because you lack out of sample values of some of the variables What s more the history of forecasting is replete with examples that work well in sample but fall apart out of sample A common compromise is to reserve part of the data by not including it in the estimation sample effectively pretending that the reserve sample is in the future Then con duct an out of sample forecast over the reserved sample taking advantage of the known values of the explanatory variables and observed outcomes We ll walk through this sort of exercise in Sample Forecast Samples One other disadvantage of in sample forecasting It s really hard to get someone to pay you to forecast
25. estimated coefficients We can now use the model to fore cast from the VAR Click set the Solution sample to 2001 2005 and hit Then do Proc Make Graph Check both Actuals and Active and set the Sample for Graph to 2000M1 2005M4 In this particular example the vec tor autoregres sion did a good job of forecasting for several periods and essentially flatlined by a year out The model taught in our introductory economics course was linear because it s hard for peo ple to solve nonlinear models Computers are generally fine with nonlinear models Hint If you prefer instead of creating a model and then copying in a VAR you can use Proc Make Model from inside the VAR to do both at once EViews Illustrated book Page 379 Monday February 25 2013 10 06 AM 380 Chapter 15 Super Models although there are some nonlinear models that are too hard for even a computer to solve For the most part though the steps we just walked through would have worked just as well for a set of nonlinear equations Rich Super Models The model object provides a rich set of facilities for everything from solving intro homework problems to solving large scale macroeconometric models We ve only been able to touch the surface To help you explore further on your own we list a few of the most prominent features Models can be nonlinear Various controls over the numerical procedures used are provided for hard problems Diagnost
26. for each series in the group So we won t discuss these further Left and Right Axes in Group Line Graphs Well truth be told there s one element of group line graphs that is worth discussing The line graph to the right from Output_and_Unemployment wf 1 shows real GDP and the civil ian unemployment rate The first thing you ll notice about this graph is that it s completely use less GDP and unemployment have different units of measure ment GDP being measured in billions of 2000 dollars and unemployment in percentage points The former scale is so much larger than the latter scale that unemployment is all but invisible EViews Illustrated book Page 140 Monday February 25 2013 10 06 AM Group Graphics 141 Since the two series have different units of measurement we need two verti cal axes so that each series can be associated with meaningful units Click the button and switch to the Axes amp Scal ing group Select series 2 Real GDP in the Series axis assignment field and click on the Right radio button Now we have a meaningful graph We can see that GDP is strongly trended while unem ployment isn t We can also see something of an inverse relation between bumps in GDP and unemployment One more option is immediately relevant Return to the Axes Scales tab and click Overlap lines cross in the Vertical axes labels field EViews Illustrated book Page 141 Monday February 25 2
27. ll trumpet the White results Compare the results here to the least squares results shown on page 335 The coefficients as well as the sum mary panel at the bottom are identi cal This reinforces the point that we re still doing a least squares esti mation but adjusting the standard errors The reported t statistics and p values reflect the adjusted standard errors Some are smaller than before and some are larger Hypothesis tests computed using Coefficient Diag nostics Wald Coefficient Restric tions correctly account for the adjusted standard errors The Omit ted Variables and Redundant Vari ables tests do not use the adjusted standard errors EViews Illustrated book Page 340 Monday February 25 2013 10 06 AM Nonlinear Least Squares 341 Nonlinear Least Squares Much as Moli re s bourgeois gentil homme was pleased to discover he had been speaking prose all his life you may be happy to know that in using the ls command you ve been doing nonlinear estimation all along Here s a very simple estimate of trend NYSE volume growth from the data set NYSEVolume wf1 ls volume c trend Switch to the Representations view and look at the section titled Estimation Equation The least squares command orders estimation of the equation except that since EViews doesn t display Greek letters C 1 is used for and C 2 is used for If you double click on you ll see that the first two e
28. ten and a moving average of order q or MA q looks like Note that the moving average error is a weighted average of the current innovation and past innovations where the autoregressive error is a weighted average of the current innovation and past errors Hint The number of correlations used in the Q statistic does not correspond to the order of serial correlation If there is first order serial correlation then the residual cor relations at all lags differ from zero although the correlation diminishes as the lag increases Convention Hint There are two sign conventions for writing out moving average errors EViews uses the convention that lagged innovations are added to the current innovation This is the usual convention in regression analysis Some texts mostly in time series analysis use the convention that lagged innovations are subtracted instead There s no consequence to the choice of one convention over the other ut r1ut 1 r2ut 2 e t ut r1ut 1 r2ut 2 rput p e t ut et vet 1 ut et v1et 1 vqet q EViews Illustrated book Page 322 Monday February 25 2013 10 06 AM Correcting for Serial Correlation 323 Conveniently the Breusch Godfrey test with q lags specified serves as a test against an MA q process as well as against an AR q process Autoregressive Regressive Moving Average ARMA Errors Aut
29. 1 et rut 1 r 0 9 r 0 yt bxt ut b y t b xt et yt bxt y t et EViews Illustrated book Page 315 Monday February 25 2013 10 06 AM 316 Chapter 13 Serial Correlation Friend or Foe are estimates of the errors so we look for serial correlation in the errors by looking for serial correlation in the residuals EViews has a variety of features for looking at residuals directly and for checking for serial correlation Exploration of these features occupies the first half of this chapter Additionally you can capture both fitted values and residuals as series that can then be investigated just like any other data The command fit seriesname stores the fitted values from the most recent estimation and the special series RESID automatically contains the residuals from the most recent estimation As an example the first command here runs a regression using data from the workfile NYSEVOLUME wf1 ls logvol c trend trend 2 d log close 1 The next commands save the fitted values and the residuals and then open a group which we changed to a line graph and then prettied up a little fit logvol_fitted series logvol_resid logvol logvol_fitted show logvol logvol_fitted logvol_resid Hint Since RESID changes after the estimation of every equation you may want to use Proc Make Residual Series to store residuals in a series which won t be acciden tally ove
30. As you know you can open a spreadsheet view of a series or a group of series to get a visual display For example the spread sheet view of GPA is shown to the right Observations appear in order Push the button to bring up the Sort Order dialog which gives you the option of sorting by either observation number or the value of GPA You can sort in either Ascending low to high or Descending high to low order By sorting according to GPA we can instantly see where the problem value is located Reminder Hint Don t forget that you can hover your cursor over points in a graph to display observation labels and values EViews Illustrated book Page 196 Monday February 25 2013 10 06 AM Describing Series Just The Facts Please 197 More generally the Sort Order dialog for groups lets you sort using up to three series to order the observations Describing Series Just The Facts Please Open a series and click the button The dropdown menu shows the tools available for looking at the series We begin with the basic descriptive statistics Stats Panel from Histogram and Stats Histograms and basic statistics are generated through the Descriptive Statistics amp Tests Histogram and Stats menu item The data used for computing descriptive statistics is as always restricted to the current sample Let s first elimi nate reported grades that are almost certainly data errors smpl if gpa gt 1 and gpa lt 5 Hint So
31. F statistics 76 213 functions data transforming 112 113 date 91 107 109 defining 92 mathematical 90 NA values and 92 naming 89 90 random number 90 91 114 statistical 114 string 104 105 G Garch graphs 353 Generalized ARCH GARCH model 353 Generalized Method of Moments GMM 347 348 genr command 86 303 geometric averages 266 267 GIF 126 global fit options 148 GMM Generalized Method of Moments 347 348 gmm command 347 graph bar 157 boxplot tab 191 color 189 fill area tab 189 multiple scatters 148 scales 141 xy line 152 graph objects 120 121 graphic files inability to import 126 graphing transformed data 147 Graphs identifying information 119 196 graphs area 132 area band 142 automatic update 120 axis borders 136 background color 182 bar 133 box plots 214 color monochrome 125 154 155 188 192 command line options 181 copying 125 correlograms 318 319 321 cumulative distribution 207 207 customizing 120 124 167 169 170 175 181 192 default settings 193 194 412 413 deleting elements 168 dot plot 132 error bar 145 146 exporting 125 126 194 412 413 forecast 146 frame tab 182 freezing 119 121 Garch 353 grouped 137 155 high low 143 144 histograms 5 197 199 203 205 kernel density 206 line 6 131 140 142 lines in 122 123 171 172 173 175 188 192 EViews Illustrated book Page 418 Monday February 25 2013 10 06 AM L 419 mi
32. February 25 2013 10 06 AM 6 Chapter 1 A Quick Walk Through Let s create a line graph to give us a visual picture of the relation between volume and time Hit the button again this time choosing Graph and selecting Line amp Symbol on the left hand side of the dia log under Graph Type EViews graphs the date on the horizontal axis and the value of VOLUME on the vertical Unsurpris ingly perhaps the most important descrip tive aspect of our data is that volume is a heck of a lot bigger than it used to be Looking at different samples We ve learned that in the early 21st cen tury NYSE volume is many orders of magnitude greater than it was at the close of the 19th In retrospect the picture isn t surprising given how much the economy has grown over this period Trading volume has grown enormously over more than a century it s grown so much in fact that numbers from the early years are barely visible on the graph Let s try a couple of different approaches to getting a clearer picture As a first pass we ll look at only the last three years or so of data To limit what we see to this period we need to change the sample Click on the button to get the dialog box shown to the right The upper field marked Sample range pairs or sample object to copy indicates that all observations are being used Replace all with the beginning and ending dates we want In this case use 2001q1 for the first da
33. If we could get an equation with some upward curvature perhaps we could do a bet ter job of matching up with the data One way to specify a curve is with a quadratic equa tion such as For this equation we need a squared time trend As in most computer programs EViews uses the caret for exponentiation In the command pane type series tsqr t 2 volume log 2 629649 0 017278t volume log volumet log a b1t b2t2 e EViews Illustrated book Page 16 Monday February 25 2013 10 06 AM Estimating your first regression in EViews 17 To see that this does give us a bit of a curve double click Then in the series window choose View Graph and select Line to see a plot showing a reassuring upward curve Close the equation window and any series windows that are clutter ing the screen Don t close the workfile window Now let s estimate a regression including to see if we can do a better job of matching the data EViews is generally quite happy to let you use a mathematical expres sion right in the LS command rather than having to first generate a variable under a new name To illustrate this capability type in the command pane ls log volume c t tsqr We ve typed in log volume instead of the series name LOGVOL and could have typed t 2 instead of tsqr thus illustrating that you can use either a series name or an algebraic expression in a regr
34. Month dd yyyy gives January 1 2005 and datestr 731946 ww produces a text string representation of the week number 6 More date formats are discussed in the User s Guide The inverse of datestr is dateval which converts a string into a date number dateval is particularly useful when you ve read in text representing dates and want a numerical version of the dates so that you can manipulate them The file SPClose Text excerpt txt has the closing prices for the S amp P 500 for the first few days of 2005 Reading this into EViews gives a numeric series for SP500CLOSE and an alpha series for CLOSE DATE The command series tradedate dateval closedate gives a numerical date series which is then available for further manipulation makedate translates ordinary numbers into date numbers for example makedate 1999 yyyy returns 729754 the first day in 1999 The format strings used for the last argument of makedate are also discussed in the User s Guide EViews Illustrated book Page 107 Monday February 25 2013 10 06 AM 108 Chapter 4 Data The Transformational Experience The workfile SPClose2005 wf1 includes S amp P 500 daily closing prices SP500CLOSE on a given YEAR MONTH and DAY To convert the last three into a usable date series use the command series tradedatenum makedate year month day yyyy mm dd To turn these into a series that looks like January 3 2005
35. Panel Estimation 279 There s a general notion from classical Solow growth theory that high population growth leads to lower per capita output conditional on available technol ogies We can test this theory by regressing gross domestic prod uct per capita relative to the United States on the rate of pop ulation growth measured as the change in the log of population Results from a simple regression seem to support such a theory At least we can say that the coefficient is statistically signifi cant In fact looking at the p value the coefficient is off the scale significant But is the effect of population growth important We can try to get a better handle on this by comparing a couple of countries let s use the Central African Republic and Canada since they re at opposite ends of the development spectrum Set the sample using smpl if isocode CAF or ISOCODE CAN Now open a window on population growth with the command show d log pop Reminder Variable names aren t case sensitive in EViews isocode and ISOCODE mean the same thing but string comparisons using are In this particular data set country identifiers have been coded in all caps CAN works can doesn t Hint in two parts The function d takes the first difference of a series and the first difference of a log is approximately the percentage change Hence d log pop gives the percentage growth o
36. The Import Business 39 Make It Slightly Easier Hint Instead of choosing the menu File Open Foreign Data as Workfile right click in any empty space inside EViews not in the command pane or inside an open window and choose Open Foreign Data as Workfile Make It Really Easy Hint Just drag and drop any data file onto an empty space inside EViews If EViews understands the data in the file the file will pop open ready to read You may have to answer a couple of questions first EViews Illustrated book Page 39 Monday February 25 2013 10 06 AM 40 Chapter 2 EViews Meet Data An Excel lent Import Source The second lowest common denomina tor file format is a Microsoft Excel spreadsheet Here s an excerpt of academic salaries by discipline xls available on the EViews website Note that variable names are conve niently provided in the first row of the file Use File Open For eign Data as Work file and point to the desired Excel file EViews does a quick analysis of the Excel file and opens the Spreadsheet read dia log EViews Illustrated book Page 40 Monday February 25 2013 10 06 AM The Import Business 41 The Spreadsheet read dialog displays lots of options but most of the time if you just hit EViews will correctly guess what you want done Note for example that EViews has fig ured out that the first line holds variable names rather than data To see what EViews is plann
37. This is called ARCH in mean or ARCH M and is added to the specification using the ARCH M drop down menu in the upper right of the Specification tab Here we ve changed the model to GARCH 1 1 and entered the variance in the structural equation EViews labels the structural coefficient of the ARCH M effect GARCH Notice that the structural effect of ARCH M is almost significant at the five percent level EViews Illustrated book Page 354 Monday February 25 2013 10 06 AM Maximum Likelihood Rolling Your Own 355 Is the 0 014 estimated GARCH M coefficient large Again we look at the conditional variance using the Garch Graph menu item In a few periods the conditional variance reaches 400 to 500 so the struc tural effect is on the order of 6 or 7 the estimated coefficient multi plied by the estimated conditional variance That s larger than sev eral of the monthly dummies But for most of the sample the GARCH M effect is relatively small Maximum Likelihood Rolling Your Own Despite EViews extensive selection of estimation techniques sometimes you want to cus tom craft your own EViews provides a framework for customized maximum likelihood esti mation mle The division of labor is that you provide a formula defining the contribution an observation makes to the likelihood function and EViews will produce estimates and the expected set of associated statistics For an example we ll return to the weighte
38. a b EViews Illustrated book Page 14 Monday February 25 2013 10 06 AM Estimating your first regression in EViews 15 Whether you use the menu or type a command EViews pops up with regression results EViews has estimated the inter cept and the slope Note that our eyeballing wasn t far off We ve estimated an equation explaining LOGVOL that reads Having seen the picture of the scatter diagram on page 13 we know this line does a decent job of summarizing over more than a century On the other hand it s not true that in each and every quarter LOGVOL equals which is what the equation suggests In some quarters volume was higher and in some the volume was lower In regression analysis the amount by which the right hand side of the equation misses the dependent variable is called the residual Calling the residual e e stands for error we can write an equation that really is valid in each and every quarter Since the residual is the part of the equation that s left over after we ve explained as much as possible with the right hand side variables one approach to getting a better fitting equa tion is to look for patterns in the residuals EViews provides several handy tools for this task which we ll talk about later in the book Let s do something really easy to start the explora tion a 2 629649 b 0 017278 LOGVOL 2 629649 0 017278t volume log 2
39. and stores it in the active workfile Objects and Commands When we type ls lnwage c ed it looks like we ve simply typed a regression command But behind the scenes EViews has created an equation object just like those we see stored in the workfile with the icon The difference is that the equation object is untitled and so not stored in the workfile Inter nally all commands are carried out as operations applied to an object Rather than give the ls command we could have issued the two commands equation aneq aneq ls lnwage c ed The first line creates a new equation object named ANEQ The second applies the ls opera tion to ANEQ The dot between ANEQ and ls is object oriented notation connecting the object aneq with a particular command ls The two lines could just as easily have been combined into one as in equation aneq ls lnwage c ed Hint Use the command pon in a program see Chapter 16 Get With the Program to make every window print or get redirected as it opens Poff undoes pon Unfortu nately pon and poff don t work from the command line only in programs EViews Illustrated book Page 398 Monday February 25 2013 10 06 AM Workfile Backups 399 So ls is actually an operation defined on the equation object Similarly output is a view of an equation object The command aneq output opens a view of ANEQ displaying the estimation output Do you need to understand the abstract con
40. button opens the VAR Specification dialog Enter the vari ables to be explained in the Endoge nous Variables field EViews Illustrated book Page 361 Monday February 25 2013 10 06 AM 362 Chapter 14 A Taste of Advanced Estimation EViews estimates least squares equations for both series Impulse response To answer the question How do the series evolve following a shock to the error term click the button The phrase following a shock is less straightforward than it sounds In general the error terms will be correlated across equations so one wants to be careful about shocking one equation but not the other And how big a shock One unit One standard deviation You control how you deal with these questions on the Impulse Definition tab For illustration purposes let s con sider a unit shock to each error term Once we choose a specification we get a set of impulse response functions We get a plot of the response over time of each endogenous variable to a shock in each equation In other words we see how G responds to shocks to both the G equation and the GV equation and sim ilarly how GV responds to shocks to the GV equation and the G equation EViews Illustrated book Page 362 Monday February 25 2013 10 06 AM Vector Autoregressions VAR 363 By default there are other options we get one figure containing all four impulse response graphs The line graph in the upper left hand corne
41. choose Alpha series EViews initializes DISCIPLINE with blank cells Move one cell to the right of DISCIPLINE and enter SALARY this time using the radio button to indicate a numeric series EViews fills out the series with NAs Hint You can name a group and store it in the workfile just as you can with a series Internally a group is a list of series names It s not a separate copy of the data A series can be a member of as many different groups as you like EViews Illustrated book Page 31 Monday February 25 2013 10 06 AM 32 Chapter 2 EViews Meet Data Click the button and type away When editing a group window the Enter key moves across the row rather than down a column This lets you enter a table of data an observation at a time rather than one variable after the other the observation at a time technique is fre quently more convenient Identity Noncrisis An important side effect of thinking of our data as being arranged in a rectangle is that each row has an observation number that identifies each observation In a group window as above or in a series window the id series is called obs and appears on the left hand side of the window Obs isn t really a series in that you can t access it or manipulate it It serves to give a name to each observation When we set up an Unstructured Undated workfile EViews just numbers the observations 1 2 3 etc In dated workfiles see below dates are used for ids r
42. cross tabulation 216 221 cubic match last conversion 249 cumulative distribution function CDF 207 cumulative distribution plot 207 cumulative distributions 207 customizing EViews 404 409 D data backing up recovering 243 399 capacity 393 contracting 261 264 266 267 conversion 244 246 251 408 copying 38 52 53 55 56 246 247 250 251 expanding 264 265 importing See importing 37 limitations 393 mixed frequency 244 251 normality of 199 pasting 38 52 53 55 56 250 251 pooling See pools series See series sources of 245 storage 410 411 types 100 105 unknown 56 57 See also dates data rectangles 24 dateadd function 108 dated irregular workfiles 36 37 dated series 34 37 datediff function 108 dates converting to text 107 108 customizing format 405 display 58 59 106 107 formats 405 408 frequency conversion and 408 functions 91 107 109 uses 58 59 datestr function 107 108 dateval function 107 day function 91 degrees of freedom 67 deleting graph elements 168 pages 243 series 303 dependent variables 14 63 65 70 deprecated commands 86 desktop publishing programs exporting to 125 display filter options 27 Display Name field 30 display type 110 distributions EViews Illustrated book Page 416 Monday February 25 2013 10 06 AM F 417 cumulative 207 normal 114 199 tdist 114 do use in programs 387 documents See workfiles down frequency conversi
43. even if you re setting up a separate workfile rather than a page as in this example If you leave the Page field blank EViews assigns Untitled or Untitled1 etc for the page name Hint The name Load Workfile Page notwithstanding this command is perfectly happy to load data from Excel text files or any of the many other formats that EViews can read It s not limited to workfiles EViews Illustrated book Page 241 Monday February 25 2013 10 06 AM 242 Chapter 9 Page After Page After Page sion of links until the section Wholesale Link Creation later in this chapter looking first at By Value to New Page or Workfile Copy Extract from Current Page copies data from the active workfile page into a new page or a standalone workfile if you prefer The command is straightforward once you get over the name This menu is accessed by clicking on the New Page tab but Current Page doesn t mean the new page attached to the tab you just clicked on it means the currently active page Just pretend that the command is named Copy Extract from Active Page Choosing By Value to New Page or Workfile opens the Workfile Copy By Value dialog which has two tabs The first tab you see Extract Spec controls what s going to be copied from the active page By default everything is copied If you like you can change the sample change which objects are copied or tell EViews to copy only a random subsample
44. original monthly data A practical consequence of this is that if you change the monthly source data perhaps because of data revisions perhaps because you discovered a typo the derived quarterly data are unaffected In contrast EViews uses the concept of a link to create a live connection between series in two pages A link is a live connection If you define a quarterly IP series as a link to the monthly IP series EViews builds up an internal connection instead of making a copy of the data Every time you use the quarterly series EViews retrieves the information from the monthly origi nal making any needed frequency conversions on the fly Any change you make to the orig inal will appear in the quarterly link as well Aesthetic hint The icon for a series link looks just like a regular series icon except it s pink If the link inside the series link is undefined or broken you ll see a pink icon with a question mark EViews Illustrated book Page 251 Monday February 25 2013 10 06 AM 252 Chapter 9 Page After Page After Page Links can be created in three ways Copy Extract from Current Page By Link to New Page creates a new page from the active page linking the new series to the originals on a wholesale basis Paste Special provides an option to paste in from the clipboard by linking instead of making copies of the original EViews series Type the command link to link in just one series W
45. over the linear model You can place most any nonlinear formula you like on the right hand side of the equation If you re lucky getting a nonlinear estimate is no harder than getting a linear estimate Sometimes though you re not lucky While EViews uses the standard closed form expres sion for finding coefficients for linear models nonlinear models require a search procedure The line in the top panel Convergence achieved after 108 iterations indicates that EViews tried 108 sets of coefficients before settling on the ones it reported Sometimes no satisfac tory estimate can be found Here are some tricks that may help Re write the formula Notice we actually estimated instead of Why The first value of TREND is zero and in this particular case EViews had difficulty raising 0 to a power This very small re write got around the problem without changing anything sub stantive Sometimes one expression of a particular functional form works when a different expression of the same function doesn t Fiddle with starting values EViews begins its search with whatever values happen to be sitting in C Sometimes a search starting at one set of values succeeds when a search at different values fails If you have good guesses as to the true values use those guesses for a starting point One way to change starting values is to double click on and edit in the usual way You can also use the param statement to change several coef
46. small compared to the standard deviation of G Looking back at our plot of out of sample forecasts versus actuals one is struck with the fact that the forecasts take wider swings than the data In the data plot that opened the chapter you can see that the volatility of currency growth was much greater in the pre War period than it was post War The HERODOTUS sample includes both periods To get an accurate estimate and thus an accurate forecast we like to use as much data as possible On the other hand we don t want to include old data if the parameters have changed We can rely on a visual inspection of the plots we ve made or we can use a more formal Chow test which confirms that a change has occurred Once again see the User s Guide If we define an alternate history with EViews Illustrated book Page 233 Monday February 25 2013 10 06 AM 234 Chapter 8 Forecasting sample turtledove 1950 2000 smpl turtledove and re estimate we find much smaller seasonal effects and a much higher in the TURTLEDOVE world than there was according to HEROTODUS Glance back at the data plot which opened the chapter It shows an enormous increase in currency holdings right before the turn of the millennium and a huge drop imme diately thereafter This means that a dynamic forecast starting in the beginning of 2000 uses an anoma lous value for lagged G a problem which is carried forward In con trast a static fore
47. solid dashed and dotted lines Actually there s one other differ ence The opening graph is stretched horizontally to make it look more dramatic See Frame amp Size in the next chapter Whenever more than one series appears on a graph the question arises as to how to visu ally distinguish one graphed line from another The two methods are to use different colors and to use different line patterns Different colors are much easier to distinguish unless your output device only shows black and white Hint Unlike many customization options is only available after a graph has been frozen EViews Illustrated book Page 122 Monday February 25 2013 10 06 AM A Simple Soup To Nuts Graphing Example 123 Click the graph window s button Initially the dialog opens to the Lines amp Sym bols section The default Pattern use is Auto choice which uses solid lines when EViews renders the graph in color and pat terned lines when EViews renders the graph in black and white Using the Auto choice default the graph appears in solid lines distinguished by colors on your display screen but patterned black lines are used if you print from EViews to a monochrome printer This default is usually the right choice But imagine that you re producing a graph for a doc ument that some readers will read electronically and therefore in color while others will read in a book and therefore in black and white Our best compromise
48. spreadsheets customizing defaults 409 views of 4 5 30 195 196 297 stacked spreadsheet view of pool 297 standard errors 66 69 308 staples 209 static forecasting 228 229 328 329 statistical significance 212 statistics by classification 201 210 histogram and 197 199 risk of error in 295 summary 5 table 201 stochastic simulations 380 storage options workfile 405 410 411 string comparisons 279 string functions 104 105 string loops 385 386 string variables 383 384 strings 102 103 structural forecasting 328 329 structural option 229 subpopulation means 213 214 summary statistics 5 SUR seemingly unrelated regressions 360 survivor plots 207 system estimation 357 361 system objects creating 357 T tab delimited text files importing from 43 45 table look ups 255 256 See also links tables comments in 395 exporting 396 opening multiple 221 saving 396 tabulation 105 199 200 216 221 tdist distribution 114 templates graph 174 177 testing by classification 212 213 Chow 233 coefficients 67 75 78 360 empirical distribution 207 fixed random effects 282 hypothesis 67 68 74 78 211 212 mean 211 median 211 series 210 213 time series 213 unit root 330 variance 211 White heteroskedasticity 339 text converting to dates 107 108 in graphs 122 169 170 175 text files importing data from 43 47 time series analysis 329 333 time series data 275 time serie
49. sum takes the low frequency number and divides it equally among the high frequency observations If first quarter sales were 600 widgets Constant match sum would set Janu ary February and March sales to 200 widgets each Quadratic match sum and Quadratic match average In 2004 GDP rose every quarter It s a little strange to assume that GDP is constant across months within a quarter and then jumps at quarter s end Quadratic conversion estimates a smooth quadratic curve using the data from the current quarter and the previous and suc ceeding quarters This curve is then used to interpolate the data within the quarter Match average forces the average of the interpolated numbers to match the original quarterly fig ure while match sum matches on the sum of the generated high frequency numbers Linear match last and Cubic match last Both Linear match last and Cubic match last begin by copying the quarterly more gener ally the low frequency source value into the last monthly observation in the corresponding quarter more generally the last corresponding high frequency date Values for the remain ing months are set by linear interpolation between final months in the quarter for Linear match last and by interpolating along natural cubic splines see the User s Guide for a defini tion for Cubic match last Down Frequency Conversions There s something slightly unsatisfying about low to high frequency conversion i
50. t any different from any other series but Pool objects let us do some nifty tricks with them The first step in pooled analysis is to give EViews a list of the suffixes CAN FRA etc that identify the countries Click on the button select New Object and choose Pool Reverse causation alert There s good reason to believe that countries becoming richer leads to lower population growth Thus there s a real issue of whether we re picking up the effect of output on population growth rather than population growth on output The issue is real but it hasn t got anything to do with illustrating the use of pools so we won t worry about it further Hint The cross section identifier needn t be placed as a suffix You can stick it any where in the series name so long as you re consistent EViews Illustrated book Page 292 Monday February 25 2013 10 06 AM Getting Your Feet Wet 293 Simply type the country identifiers one per line into the blank area and then name the pool by clicking on the button In our example the window ends up looking as shown to the right Click on the button in the Pool window For this first example enter Y in the Dependent variable field and C D LOG POP in the Common coefficients field In a pooled analysis the in the variable names gets replaced with the ids listed in the pool object Clicking gives us a regression that s just like the regres
51. tistic and p value have the usual interpretations In our equation there is very strong evidence that the serial correlation coefficient doesn t equal zero confirming all our earlier sta tistical tests ut r1ut 1 r2ut 2 e t vet 1 r EViews Illustrated book Page 323 Monday February 25 2013 10 06 AM 324 Chapter 13 Serial Correlation Friend or Foe Now let s look at the top panel where we see that the number of observations has fallen from 430 to 429 EViews uses observations from before the start of the sample period to esti mate AR and MA models If the current sample is already at the earliest available observa tion EViews will adjust the sample used for the equation in order to free up the pre sample observations it needs There s an important change in the bottom panel too but it s a change that isn t explicitly labeled The summary statistics at the bottom are now based on the innovations rather than the error For example the gives the explained fraction of the variance of the dependent variable including credit for the part explained by the autoregressive term Similarly the Durbin Watson is now a test for remaining serial correlation after first order serial correlation has been corrected for Serial Correlation and Misspecification Econometric theory tells us that if the original equation was otherwise well specified then correcting for serial correlatio
52. 365 Looking At Model Solutions 370 More Model Information 372 Your Second Homework 374 Simulating VARs 378 Rich Super Models 380 Quick Review 380 CHAPTER 16 GET WITH THE PROGRAM 381 I Want To Do It Over and Over Again 381 You Want To Have An Argument 382 Program Variables 383 Loopy 385 Other Program Con
53. 4 Data The Transformational Experience Hint Should one use a common sample or not EViews requires a choice in several of the procedures we re looking at in this chapter The most common practice is to find a common sample to use for the entire analysis That way you know that different answers from different parts of the analysis reflect real differences rather than different inputs On the other hand restricting the analysis to a common sample can mean ignoring a lot of data So there s no absolute right or wrong answer It s a judgment call EViews Illustrated book Page 216 Monday February 25 2013 10 06 AM Describing Groups Just the Facts Putting It Together 217 Choosing the N Way Tabulation view and accepting the defaults in the Crosstabulation dialog produces the output shown to the right Let s begin with the table appearing at the bottom Table Facts The first column of the table gives counts for high test scoring applicants 5 had top grades 552 had high but not top grades and 306 had low grades for a total of 863 applicants with high i e above average test scores Reading across the first row 5 of the students with top grades had high LSATs 5 were below average so there was a total of 10 students with top grades The bottom row reports the totals for each column and the right most column reports the totals for each row The Total Total bottom right is the number
54. 42 1 wfcreate m december 1941 january 1942 Canadians and Americans among others write dates in the order month day year Out of the box EViews comes set up to follow this convention You can change to the European convention of day month year by using the Options Dates amp Frequency Conversion menu You can also switch between the colon and frequency delimiter e g 41 12 versus 41m12 Hint If your date string includes spaces put it in quotes Hint Use frequency delimiters rather than the colon 41q2 always means the second quarter of 1941 while 41 2 means the second quarter of 1941 when used in a quar terly workfile but means February 1941 in a monthly workfile Ambiguity is not your friend EViews Illustrated book Page 58 Monday February 25 2013 10 06 AM Appendix Having A Good Time With Your Date 59 The most common use of dates as data is as the id series that appears under the Obs column in spreadsheet views and on the horizontal axis in many graphs But nothing stops you from treating the values in any EViews series as dates For example one series might give the date a stock was bought and another series might give the date the same stock was sold Internally EViews stores dates as date numbers the number of days since January 1 0001AD according to the Gregorian proleptic don t ask calendar For example the series DATE created with the command series date date
55. 629649 0 017278t LOGVOL 2 629649 0 017278t e EViews Illustrated book Page 15 Monday February 25 2013 10 06 AM 16 Chapter 1 A Quick Walk Through Just as there are multiple ways to view series and groups equations also come with a variety of built in views In the equation window choose the button and pick Actual Fitted Residual Actual Fitted Residual Graph The view shifts from numbers to a picture There are lots of details on this chart Notice the two different vertical axes marked on both the left and right sides of the graph and the three different series that appear The horizontal axis shows the date The actual values of the left hand side variable called Actual and the values predicted by the right hand side called Fitted appear in the upper part of the graph In other words the thick upper line marked Actual is and the straight line marked Fitted is Actual and Fitted are plotted along the vertical axis marked on the right side of the graph fitted values rising roughly from 2 to 8 Residual plotted in the lower portion of the graph uses the legend on the left hand vertical axis Whether we look at the top or bottom we see that the fitted line goes smack though the mid dle of in the early part of the sample but then the fitted value is above the actual data from about 1930 through 1980 and then too low again in the last years of the sample
56. Ends Saving Tables and Almost Tables It s nice to look at output on the screen but eventually you ll probably want to transfer some of your results into a word pro cessor or other program One fine method is copy and paste You can also save any table as a disk file through Proc Save table to disk or the Save table to disk context menu item when you right click in the table Either way you get a nice list of choices for the table format on disk If you re planning on reading the table into a spreadsheet or database program choose Comma Separated Value or Tab Delimited Text ASCII If the table s eventual destination is a word processor you can use Rich Text Format to preserve the formatting that EViews has built into the table Saving Graphs and Almost Graphs Saving graphs works much like saving tables In addition to using copy and paste to transfer your graph into another program you can save any graph as a disk file through Proc Save graph to disk or the Save graph to disk context menu item when you right click in the graph Sim ply select one of the available graph disk formats Hint If you re looking at a view of an object that would become a table if you froze it regression output or a spreadsheet view of a series are examples Save table to disk shows up on the right click menu even though it isn t available from Proc Hint If you re looking at a view of an object that would become a graph if
57. February 25 2013 10 06 AM 284 Chapter 11 Panel What s My Line Looking at all possible cross sec tions isn t always very informa tive This is one case where too much data means too little infor mation Combined cross sec tions works much better when there are only a small number of countries If we limit the sample to the Central African Republic and Canada as we did above the plot is much more informa tive Going through the same process we can choose Individual cross sections to get a different picture of the same data Where the previous graph visually emphasized the differ ence in income levels between the Central Afri can Republic and Canada this graph is better at showing the relative trends over time More Panel Estimation Techniques We ve merely brushed the surface of the panel estimation techniques that EViews provides discussing fixed effects models and a couple of panel special graphing techniques The EViews User s Guide devotes an entire chapter to panel estimation Here are a few items you may want to check out Hint EViews doesn t produce a legend for this sort of graph by default You can dou ble click on the graph go to the legend portion of the dialog and tell it do so EViews Illustrated book Page 284 Monday February 25 2013 10 06 AM One Dimensional Two Dimensional Panels 285 Fixed effects in the time dimension or in time and cross section simultaneously
58. Fixed effect estimation puts in an intercept for every country and changes slightly how the results are reported The intercept is now reported in two parts Nothing else in the report changes The line marked C reports the average value of the intercept for all the countries in the sample The lines marked for the individual countries give the country s intercept as a deviation from that overall average In this example the overall average intercept is 76 and the intercept for Canada is 96 20 above 76 Testing Fixed Effects Fixed effects specifications are common enough that EViews builds in a test for country specific intercepts against a single com mon intercept After a pooled esti mate specifying fixed effects choose and then the menu Fixed Random Effects Test ing Redundant Fixed Effects Likelihood Ratio Both F and tests appear at the top of the view Since the hypothesis of a common intercept is wildly rejected there s more to the fixed effect specifica tion than just that it gives results that we like x2 EViews Illustrated book Page 296 Monday February 25 2013 10 06 AM Playing in the Pool Data 297 Playing in the Pool Data Pooled series are just plain old series that share a naming convention All the usual opera tions on series work as expected But there are some extra features so that you can examine or manipulate all the series in a pool in one operation Spreadsheet Views
59. Give A Graph A Fair Break 178 Options Options Options 181 The Impact of Globalization on Intimate Graphic Activity 193 Quick Review 194 CHAPTER 7 LOOK AT YOUR DATA 195 Sorting Things Out 196 Describing Series Just The Facts Please 197 EViews Illustrated book Page ii Monday February 25 2013 10 06 AM Table of Contents iii Describing Series Picturing the Distribution 203 Tests On Series 210 Describing Groups Just the Facts Putting It Together 214 CHAPTER 8 FORECASTING
60. Page 254 Monday February 25 2013 10 06 AM Have A Match 255 Have A Match One key to thinking about which data should be collected in a single EViews page is that all the series in a page share a common identifier One page might hold quarterly series where the identifier is the date Another page might hold information about U S states where the identifier might be the state name or just the numbers 1 through 50 What you can t have is one page where some series are identified by date and others are identified by state So far the examples in this chapter have all used dates for identifiers Because EViews has a deep understanding of the calendar it knows how to make frequency conversions for example translating monthly data to quarterly data So while data of different frequencies needs to be held in different pages linking between pages is straightforward What do you do when your data series don t all have an identifier in common EViews pro vides a two step procedure Bring all the data which does share a common identifier into a page creating as many different pages as there are identifiers Use match merge to connect data across pages The workfile Infant Mortality Rate wf1 holds two pages with data by state The page Mor tality contains infant mortality rates and the page Revenue contains per capita revenue Is there a connection between the two Excerpts of the data look like There
61. R2 EViews Illustrated book Page 347 Monday February 25 2013 10 06 AM 348 Chapter 14 A Taste of Advanced Estimation The resulting estimate is close to the 2SLS estimate but it s not identical By default EViews applies one of the many available options for estimation that is robust to heteroskedasticity and serial correlation Clicking the button reveals the GMM Specification tab The entire right side of this tab is devoted to the choice of robust esti mation methods See the User s Guide for more information Orthogonality Conditions The basic notion behind GMM is that each of the instruments is orthogonal to a specified function You can specify the function in any of three ways If you give the usual dependent variable followed by independent variables series list the function is the residual If you give an explicit equation linear or nonlinear the function is the value to the left of the equal sign minus the value to the right of the equal sign If you give a formula with no equal sign the formula is the function See System Estimation below for a brief discussion of GMM estimation for systems of equa tions EViews Illustrated book Page 348 Monday February 25 2013 10 06 AM Limited Dependent Variables 349 Limited Dependent Variables Suppose we re interested in studying the determinants of union membership and that coincidentally we have data on a cross section of work
62. Random effect models A variety of procedures for estimating coefficient covariances A variety of panel oriented GLS generalized least squares techniques Dynamic panel data estimation Panel unit root tests One Dimensional Two Dimensional Panels Panels are designed for data that s inherently two dimensional Effectively data are grouped by both cross section and time period Grouping data is sometimes useful even when there is only one dimension to group along In other words sometimes it s useful to pretend that data comes in a panel even when it doesn t In particular this can be a useful trick for esti mating separate intercepts for each group Information on wages in logs education in years age race and state was collected as part of the Current Population Survey CPS in March 2004 The workfile CPSMAR2004extract wf1 con tains an extract with usable data for about 100 000 individuals We can use this data to test the theory that Asians earn more than others after accounting for observable differences such as education and age A standard method for look ing at this kind of question is to regress wages on education age and a dummy variable for being Asian We then ask whether the Asian effect is positive Using the CPS data we get the results shown here EViews Illustrated book Page 285 Monday February 25 2013 10 06 AM 286 Chapter 11 Panel What s My Line Our
63. Really Different 261 EViews has pasted data into the ByState page using what s called a match merge In this case we ve gotten the obvious and desired result The ED series in the ByState page gives the average number of years of education in each state the UNION series gives the percent unionized etc Let s back up and talk separately about the match step and the merge step The match step connects the identifiers across pages In this case we want to connect obser vations for individuals and observations for states according to whether they have the same state GMSTCEN value The merge step maps the large number of observations for individuals into a single value for each state in the ByState page The default contraction method Mean is to average the val ues for individuals within a state Fib Warning In fact we haven t gotten the desired result for a subtle reason involving the sample Finding the error lets us explore some more features in the next section Contracted Data For the moment we ll pretend everything is okay Ex post obvious hint When you average a 0 1 variable like UNION you get the frac tion coded as a 1 That s because adding up 0 1 observations is the same as count ing the number of 1 s So taking the average counts the number of 1 s and then divides by the number of observations EViews Illustrated book Page 261 Monday February 25 2013 10 06 AM 262 Chapte
64. Street data the Daily 7 day for keeping track of graduate student work hours EViews Illustrated book Page 36 Monday February 25 2013 10 06 AM The Import Business 37 You can t create a dated irregular structure from the Workfile Create dialog Instead you read the data into one of the available structures and then restructure the workfile For example the workfile Russell3000Regular wf1 holds Daily 5 day week data A small excerpt is shown to the right To change this to Dated irregu lar double click on Range in the workfile window or choose Proc Structure Resize Current Page to bring up the Workfile structure dialog Pick Dated specified by date series Enter the name of the series containing observation dates in the Identi fier series field Notice that after restructuring November 26 1987 which was previously shown as NA has disappeared from the data set The Import Business Once you ve created an empty workfile you can turn to filling it up with your data Hint Date functions work as expected in Dated irregular workfiles However lags pick up the preceding observation as in unstructured workfiles not the preceding date as in regular dated workfiles In our original file one lag of 11 27 1987 was 11 26 1987 which happened to be NA In our new file one lag of 11 27 1987 is 11 25 1987 EViews Illustrated book Page 37 Monday February 25 2013 10 06 AM 38 Chapter 2 EV
65. Table 2 so as to make it easy for you to match up the displayed numbers with the EViews output From now on we ll be better behaved volumet log 2 629649 0 089576 0 017278 t 0 000334 ser 0 967362 R2 0 852357 t2 R2 EViews Illustrated book Page 65 Monday February 25 2013 10 06 AM 66 Chapter 3 Getting the Most from Least Squares The most important elements of EViews regression output are the estimated regression coef ficients and the statistics associated with each coefficient We begin by linking up the num bers in the inline display or equivalently column 1 of Table 2 with the EViews output shown earlier The names of the independent variables in the regression appear in the first column labeled Variable in the EViews output with the estimated regression coefficients appearing one column over to the right labeled Coefficient In econometrics texts regression coeffi cients are commonly denoted with a Greek letter such as or or occasionally with a Roman In contrast EViews presents you with the variable names for example TREND rather than The third EViews column labeled Std Error gives the standard error associated with each regression coefficient In the scientific reporting displays above we ve reported the standard error in parentheses directly below the associated coefficient The standard error is a measure of uncert
66. The User s Guide describes several advanced statistical tools which can be used with panels EViews Illustrated book Page 289 Monday February 25 2013 10 06 AM 290 Chapter 11 Panel What s My Line EViews Illustrated book Page 290 Monday February 25 2013 10 06 AM Chapter 12 Everyone Into the Pool Suppose we want to know the effect of population growth on output We might take Cana dian output and regress it on Canadian population growth Or we might take output in Grand Fenwick and regress it on Fenwickian population growth Better yet we can pool the data for Canada and Grand Fenwick in one combined regression More data better esti mates Of course we ll want to check that the relationship between output and population growth is the same in the two countries before we accept combined results Pooling data in this way is so useful that EViews has a special facility the pool object to make it easy to work with pooled data We begin this chapter with an illustration of using EViews pools Then we ll look at some slightly fancy arrangements for handling pooled data Getting Your Feet Wet The file PWT61PoolExtract wf1 available from the EViews website contains annual data on popu lation and output relative to the United States extracted from the Penn World Tables for the G7 countries Canada France Germany Great Britain Italy Japan and the United States The first thing you ll notice
67. The tab Page Destination lets you give the about to be created page a name of your choice This is also the spot for tell ing EViews whether you want a new workfile or a page in the existing workfile EViews Illustrated book Page 242 Monday February 25 2013 10 06 AM Renaming Deleting and Saving Pages 243 If all you want to do is copy the contents of the existing page into a new page simply drag ging the page tab and dropping it on the New Page tab of any workfile A plus sign will appear when your cursor is over an appropriate area Renaming Deleting and Saving Pages To rename delete or save a page right click on the page tab to bring up a context menu with choices to let you no surprises here rename delete or save the page The only mild subtlety is that saving a page actually saves an entire workfile containing that page as its only contents You can also save a page as a workfile on the disk by using the Proc Save Current Page menu Messing up hint EViews doesn t have an Undo function When you re about to make a bunch of changes to your data and you d like to leave yourself a way to back out consider using Copy Extract from Current Page to make a copy of the active page Then make the trial changes on the copy If things don t work out you still have the original data unharmed on the source page Hint Save Workfile Page will write Excel files text files etc as well as EViews workfil
68. Then open a group window which will look more or less like the second window above Select all the data by dragging the mouse Then copy to the clipboard Click in All Data wf1 and open a group with all the series just as we did for Last Three Hit highlight the last three rows and paste The data copied from Last Three replaced the NAs and we re done The same principle works with more than two data sources Make your all data workfile big enough to hold all your data and then copy into the appropriate rows from each smaller workfile sequentially Hint SaveAs before changing range This way an error doesn t mess up the original version of First Two wf1 Hint Be careful that each group has all the series in the same order EViews is just copying a rectangle of numbers If you accidentally change the order of the series EViews will accidentally scramble the data EViews Illustrated book Page 53 Monday February 25 2013 10 06 AM 54 Chapter 2 EViews Meet Data Expanding from the middle out EViews expands a workfile by adding space for new observations at the end of the workfile range Adding observations in the middle requires two steps First expand the workfile just as we ve done above Second move observations down to the new bottom of the rectangle To accom plish the latter open a group including all series in the workfile select the rows that should go to the bottom and hi
69. We would have ended up with an entire series of NAs Our original use of smpl statements to avoided propagating NAs by having the first lagged value be the value of the first observation which was 0 as we intended Functions Are Where It s EViews function names mostly begin with the symbol There are a lot of functions which are documented in the Command and Programming Reference We ll work through some of the more interesting ones below in the section Relative Exotica Here we look at the basics Several of the most often used functions have reserved names meaning these functions don t need the sign and that the function names cannot be used as names for your data Moral When you use lagged variables in an equation think carefully about whether the lags are picking up the observations you intend EViews Illustrated book Page 89 Monday February 25 2013 10 06 AM 90 Chapter 4 Data The Transformational Experience series Don t worry if you accidentally specify a reserved name EViews will squawk loudly To create a variable which is the logarithm of X type series lnx log x Other functions common enough that the sign isn t needed include abs x for abso lute value exp x for and d x for the first difference i e The func tion sqr x means not for what are BASICally historical reasons for squares just use 2 EViews provides the expected pile of
70. also clicked the notched and pro portional to obser vations radio buttons in the dia log The distribu tion of grades is higher for non Washington residents but the ranges for residents and non residents mostly overlap Statistics don t lie but they can mislead In this plot the upper staple for non residents is higher than the staple for residents A little investigation shows that while GPA is measured on a four point scale one observation for a non residents was recorded 4 1 Because some schools use something other than a four point scale we don t know if this GPA is an error or not But we probably shouldn t conclude anything about the difference between in and out of state applicants based on this one data point Tests On Series Up to this point we ve been looking at ways to summarize the data in a series Now we move on to formal hypothesis tests The tests corresponding to the descriptive statistics we ve looked at are found under the Descriptive Statistics amp Tests menu As an example having found the mean LSAT in our applicant pool we might want to test whether it differs from the national average EViews Illustrated book Page 210 Monday February 25 2013 10 06 AM Tests On Series 211 Simple Hypothesis Tests Nationally the average LSAT score is about 152 Looking at the data for University of Washington applicants we see their average was higher just over 158 It would be interesting to
71. aren t going to want to set the same options over and over and over and over Templates help some You can also set global default options through the menu Options Graphic Defaults which brings up a Graph Options dialog with essentially the same sections we ve seen already Changes made here become the ini tial settings for all future graphs EViews Illustrated book Page 193 Monday February 25 2013 10 06 AM 194 Chapter 6 Intimacy With Graphic Objects Exporting The global Graph Options dialog has an extra section Exporting Make changes here to control the defaults for exporting graph ics to other pro grams Quick Review Basically EViews has a gazillion options for getting up close and personal with graphics Fortunately you rarely need this level of detail because EViews has sound artistic sensibili ties Nonetheless when there s a customization you do need it s probably available EViews Illustrated book Page 194 Monday February 25 2013 10 06 AM Chapter 7 Look At Your Data Data description precedes data analysis Failure to carefully examine your data can lead to what experienced statisticians describe with the phrase a boo boo True story I was involved in a project to analyze admissions data from the University of Washington law school An extract of the data UWLaw98 wf1 can be found on the EViews website Some of my early results were really really strange After hours of f
72. base 10 logarithms Since econometricians should never have any use for base 10 logarithms we avoid this aesthetically displeasing notation Hint If you insist on using base 10 logarithms use the log10 function And for the rebels amongst us there s even a logx function for arbitrary base logarithms ex d x x x 1 x x2 EViews Illustrated book Page 90 Monday February 25 2013 10 06 AM Your Basic Elementary Algebra 91 What if you want uniform random numbers distributed between limits other than 0 and 1 or normals with mean and variance different from 0 and 1 There s a simple trick for each of these If x is distributed uniform 0 1 then is distributed uniform a b If x is distributed N 0 1 then is distributed The corresponding EViews com mands using 2 and 4 5 for and and 3 and 5 for and are series x 2 4 5 2 rnd series x 3 5 nrnd Trends and Dates The function trend generates the sequence 0 1 2 3 You can supply an optional date argument in which case the trend is adjusted to equal zero on the specified date The results of trend 1979 appear to the right The functions year quarter month and day return the year quarter month and day of the month respectively for each observation weekday returns 1 through 7 where Mon day is 1 For instance a dummy 0 1 variable marking the postwar period could be coded series postwar year gt 1945 or a dummy variable
73. between data fields It s common because it s natural to us human types to find spaces between words At least for most modern western languages So this seems an ideal way to arrange data for the computer to read and for the most part it works fine But consider the line Is the value for the second variable 30500 or is it Sciences We know the intended answer is the former because we understand the context But if you tell EViews that your data are separated by spaces it s going to believe you and there is a space between Life and Sciences If you have complete control of the text file either because you create it or because you can edit it by hand you can mark off a single text string by placing it between quotes For exam ple put Life Sciences between quotes EViews will treat quoted material as one long string which is what we want in this case The inverse problem happens with space delimited data when a data item is missing If a column is left blank for a particular observation we humans assume there is a number missing EViews will just see the blank space as part of the delimiter So while we understand that the text excerpt to the right should be interpreted as 1 2 3 for the first observation and 4 NA 6 for the second observation all EViews sees is a long white space between the 4 and the 6 and interprets the data as 1 2 3 followed by 4 6 NA If the data ar
74. care about capitalization of names NONACADSAL and nonacadsal are the same thing EViews Illustrated book Page 28 Monday February 25 2013 10 06 AM Time to Type 29 We re ready to type numbers But there s a trick to entering your data To protect against accidents EViews locks the window so that it can t be edited To unlock click on the button Unlocked windows as shown below for example have an edit field just below the button bar One way to know that a window is locked against editing is to observe the absence of the edit field Alternatively if you start typing and nothing happens you ll remember that you meant to click on the button at least that s what usually hap pens to the author Initially all the entries in the window are NA for not available Click on the cell just to the right of the and type the first data point for nonacademic salaries 40005 Hit Enter to complete the entry Enter the rest of the data displayed at the beginning of the chapter You can use all the usual arrow and tab keys as well as the mouse to move around In addition when a cell is selected you can edit its contents in the edit field in the upper left of the window Hint Naming a series or other object enters it in the workfile at the same time it attaches a moniker In contrast Untitled windows are not kept in the workfile If you close an Untitled window Poof It s gone The key to remember is that named objects are
75. defined 95 effect on data 9 freezing 99 selecting 6 8 9 specifying 95 98 99 100 saving graphs 126 396 pages 243 programs 381 tables 396 unnamed objects 407 scatter graphs 13 146 147 See also XY line graphs scatter plot 128 scenarios model 375 376 seasonal graphs 134 136 seemingly unrelated regressions SUR 360 serial correlation causes 317 correcting for 323 326 defined 315 first order model 315 forecasting 327 329 higher order 322 misspecification 324 moving average 322 323 statistical checks for 73 319 322 visual checks for 317 319 series columnar representation 58 correlating 215 creating 8 10 27 28 85 113 dated 34 37 empty fields in 31 32 graphing multiple 148 in groups 30 32 214 221 icon indicating 4 identifying 32 34 labeling 30 31 labelling 31 renaming 48 selecting multiple 118 stack in single graph 140 temporary 113 terminology 24 tests on 210 213 viewing 4 6 11 13 30 in workfiles 27 EViews Illustrated book Page 422 Monday February 25 2013 10 06 AM U 423 See also auto series series command 28 100 254 shading in graphs 171 172 175 shocks calculating response to 362 363 significance statistical 212 simulations stochastic 380 single variable control problems 380 smpl command 96 97 sorting 196 197 space delimited text files importing from 45 spaces in auto series 94 special expressions 90 spellchecking 105 spike graphs 134
76. dif ference between the confidence intervals with and without coefficient uncertainty is all but invisible to the eye That s one reason people often don t bother including coefficient uncertainty For the record here s the command that produced the plot before we added the title and tidied it up plot ghein_dyn ghein_dyn 1 96 ghein_dyn_se_all ghein_dyn 1 96 ghein_dyn_se_nocoef ghein_dyn 1 96 ghein_dyn_se_all ghein_dyn 1 96 ghein_dyn_se_nocoef Forecast Evaluation EViews provides two built in tools to help with forecast evalua tion the Output field checkboxes Forecast graph and Forecast evaluation Hint This plot shows two forecasts and two associated sets of confidence intervals Usually you want to see a single forecast and its confidence intervals EViews can do that graph automatically as we ll see next EViews Illustrated book Page 232 Monday February 25 2013 10 06 AM Forecast Evaluation 233 The Forecast graph option auto mates the 95 confidence inter val plot The Forecast evaluation option generates a small table with a variety of statistics for comparing fore cast and actual values The Root Mean Squared Error or RMSE is the standard deviation of the fore cast errors See the User s Guide for explanations of the other statistics Our forecasts aren t bad but the confidence intervals shown in the graph above are fairly wide given the observed movement of G Similarly the RMSE is not
77. displayed in the workfile window with a variant of the ordinary series icon Typographic hint EViews thinks series are separated by spaces This means that when using an auto series it s important that there not be any spaces unless the auto series is enclosed in parentheses In our example EViews interprets y mean y as the deviation of Y from its mean as we intended If we had left a space before the minus sign EViews would have thought we wanted Y to be the dependent variable and the first independent variable to be the negative of the mean of Y EViews Illustrated book Page 94 Monday February 25 2013 10 06 AM Simple Sample Says 95 To understand named auto series it helps to know what EViews is doing under the hood For an ordinary series EViews computes the values of the series and stores them in the workfile For a named auto series EViews stores the formula you provide Whenever the auto series is referenced EViews recalculates its values on the fly A minor advantage of the auto series is that it saves storage since the values are computed as needed rather than always taking up room in memory The major advantage of named auto series is that the values automatically update to reflect changes in values of series used in the formula In the example above if any of the data in Y changes the values in Y_LESS_MEAN change automatically When we get to Chapter 8 Forecasting we ll learn about a special role that a
78. encapsulated 125 presentation techniques 64 65 printing to file 398 graphs 124 tables 221 programs 381 388 p values 68 69 76 Q Q statistic 321 322 quadratic match average conversion 249 quadratic match sum conversion 249 quantile plots 207 quarter function 91 question mark in series names 297 312 quotes in strings 108 EViews Illustrated book Page 421 Monday February 25 2013 10 06 AM 422 Index R racial groupings 286 random estimation 308 random number generators 90 91 114 range 27 recessions graphing 173 174 175 recode command 57 91 recovering data 243 399 regression analysis 13 17 auto series in 93 94 coefficients 66 commands 17 distribution 199 estimates 317 least squares 275 276 letter C in 64 74 lines 13 multiple 73 74 purpose 61 results 64 65 67 73 rolling 387 scatter plots with 147 seemingly unrelated SUR 360 tools See forecasting heteroskedasticity serial correlation VARs 361 364 378 379 Remarks field 30 Representations View 78 reserved names 366 residuals 15 17 27 78 81 304 306 315 318 restructuring workfiles 33 right hand side variable 65 RMSE root mean squared error 233 rnd function 90 114 rndseed function 90 rolling forecasts 388 rolling regressions 387 root mean squared error RMSE 233 rotation 155 S salary example academic 23 24 sample command 98 sample field 27 sample programs 381 samples common 215 216
79. error of the regression if you choose One standard deviation or by a value of your own choosing Here s the plot for our AR 1 model set to User specified 1 0 Even after two years 28 percent of the final residual will remain in the forecast Forecasting When we discussed forecasting in Chapter 8 Forecasting we put off the discussion of forecasting with ARMA errors because we hadn t yet discussed ARMA errors Now we have The intuition about forecasting with ARMA errors is straightforward Since errors persist in part from one period to the next forecasts of the left hand side variable can be improved by including a forecast of the error term EViews makes it very easy to include the contribution of ARMA errors in 0 9eT 0 81eT 0 729eT EViews Illustrated book Page 327 Monday February 25 2013 10 06 AM 328 Chapter 13 Serial Correlation Friend or Foe forecasts Once you push the button all you have to do is not do anything in other words the default procedure is to include ARMA errors in the forecast Just don t check the Structural ignore ARMA checkbox Static Versus Dynamic Forecasting With ARMA errors When you include ARMA errors in your forecast you still need to decide between static and dynamic forecasting The difference is best illustrated with an example We have data on NYSE volume through the first quarter of 2004 Let s forecast volume for the last eight quar
80. etc use the command alpha tradedate datestr tradedatenum Month dd yyyy In this command the four quotes in a row are interpreted as a quote opening a string first quote two quotes in a row which stand for one quote inside the string middle two quotes and a quote closing the string last quote It takes all four quotes to get one quote embedded in the string Date Manipulation Functions Since dates are measured in days you can add or subtract days by unsurprisingly adding or subtracting For example the function now gives the number representing the current day and time so now 1 is the same time tomorrow The functions dateadd date1 offset units_string and datediff date1 offset units_string add and subtract dates allowing for dif ferent units of time The units_string argument specifies whether you want arithmetic to be done in days dd weeks ww etc See the User s Guide for information on etc dateadd now 1 dd is this time tomorrow while dateadd date 1 ww is this time on the same day next week Suppose we want to compute annualized returns for our 2005 S amp P closing price data We can compute the percentage change in price from one observation to the next with series pct_change sp500close sp500close 1 sp500close 1 or with the equivalent built in command pch series pct_change pch sp500close To annualize a daily return we could mu
81. extra information showing the auxil iary regression used to create the test statistics reported at the top This extra regression is sometimes interesting but you don t need it for conducting the test Ljung Box Q statistic A different approach to checking for serial correlation is to plot the correlation of the resid ual with the residual lagged once the residual with the residual lagged twice and so on As we saw above in The Correlogram this plot is called the correlogram of the residuals If there is no serial correlation then correlations should all be zero except for random fluctuation To see the correlogram choose Residual Diagnostics Correlogram Q statistics from the menu A small dialog pops open allowing you to specify the number of correlations to show The correlogram for the residuals from our volume equation is repeated to the right The column headed Q Stat gives the Ljung Box Q statistic which tests for a particular row the hypothesis that all the correlations up to and including that row equal zero The column marked Prob gives the corresponding p value Continuing along with the example the Q statistic against the hypothesis that both the first and second correlation equal zero is 553 59 The probability of getting this statistic by chance is zero to three decimal places So for this equation the Ljung Box Q statistic agrees with the evi x2 EViews Illustrated book Page 321 Monday Februa
82. for a particular date a point forecast There is always some degree of uncertainty around this point forecast Assuming our model is correctly specified such uncertainty derives from two sources coeffi cient uncertainty and error uncertainty Our forecast for date is while the actual value of the series we re forecasting will be The forecast error will be The term in square brackets is the source of coeffi cient uncertainty The error term at the forecast date causes error uncertainty If you enter a name next to S E optional in the Forecast dialog EViews will save the standard error of the forecast distribution in a series It s not unusual to ignore coefficient uncertainty in evaluating a forecast If you want to exclude the effect of coefficient uncer tainty uncheck Coef uncertainty in S E calc in the dialog t y t a b xt yt a bxt ut yt y t a a b b xt ut ut EViews Illustrated book Page 231 Monday February 25 2013 10 06 AM 232 Chapter 8 Forecasting We ve stored the standard error for our dynamic forecast including coefficient uncer tainty as GHEIN_DYN_SE_ALL and analogously without coef ficient uncertainty under GHEIN_DYN_SE_NOCOEF Here s a plot of the forecast value and confidence bands measured as the forecast minus 1 96 standard errors through the forecast plus 1 96 standard errors You can see that the
83. for the arrow none filled arrow hollow arrow or perpen dicular line You can also set the arrow color width and line properties To apply the settings to all arrows in the graph check the box at the bottom of the dialog If you d like to delete an arrow select it and press the Delete key just as with other EViews objects Your Data Another Sorta Way You can sort a spreadsheet view See Sorting Things Out in Chapter 7 Look At Your Data You can also sort the data in a frozen graph Choose Sort from the button or the right click menu In addition to being able to sort according to the values in as many as three of your data series you can sort according the value of observation labels Give A Graph A Fair Break If there s a break in your sample how would you like that break to be displayed in a graph EViews offers three options in the Graph Options dialog In order to have something simple for an illustration we ve created a series 1 2 3 4 5 and then set the sample with smpl 1 2 4 5 so the middle observation is missing Hint As mentioned earlier you must freeze a graph before you sort it Sort re arranges the data so if you try to sort a histogram or other plot where sorting isn t sen sible EViews sensibly doesn t do anything EViews Illustrated book Page 178 Monday February 25 2013 10 06 AM Give A Graph A Fair Break 179 Drop excluded obs deletes the missing part of the
84. forecast in EViews We told you the mechanics were easy Theory of Forecasting Let s review a little forecasting theory and then see where EViews fits in There are three steps to making an accurate forecast 1 Formulate a sound model for the variable of interest 2 Estimate the parameters attached to the explanatory variables 3 Apply the estimated parameters to the values of the explanatory variables for the fore cast period Let s call the variable we re trying to forecast Suppose that a good way to explain is with the variable and that we ve decided on the model where in our opening example would represent the time trend lagged currency growth etc Choosing a form for the model was Step 1 Now gather data on for periods and use one of EViews myriad estimation techniques perhaps least squares but perhaps another method to assign numerical estimates and to the parameters in the model You can see the estimated parameters in the equation EQ01 above That was Step 2 Finally let s say that we want to forecast at all the dates starting at and continuing through This is the forecast sample which is to say the entry marked Forecast sample in the lower left part of the Forecast dialog Our forecast will be y y x yt a bxt ut x yt and xt t 1 T a b yt t TF TL EViews Illustrated book Page 225 Monday February 25 2013 10 06 AM 22
85. highly significant the standard errors are larger than before That s what we would expect from the much smaller 99 991 versus 51 observations sample In the state by state regression the interpretation of the union coefficient has changed Because UNION is measured as a fraction the regression now tells us that for each one percentage point increase in the unionization rate the average wage rises two and a half percent Expanded Data In order to separate out the effect of the average unionization rate from the effect of individ ual union membership we need to include both variables in our individual level regression To accomplish this we need to expand the 51 state by state observations on unionization back into the individual page linking each individual to the average unionization rate in her state of residence Econometric caution We re assuming that unionization drives wage rates Maybe Or maybe it s been easier for unions to survive in high income states The latter interpre tation would mean that our regression results aren t causal EViews Illustrated book Page 264 Monday February 25 2013 10 06 AM Expanded Data 265 To expand the data we make a link going in the other direction Copy UNION from the ByState page and Paste Special into the Cps page We ll change the name of the pasted variable to AV_UNION in order to avoid any confusion with the indi vidual union variable The first few obse
86. is to use color and line patterns Readers of the electronic version will see the color and readers in traditional media will be able to distinguish the lines by their patterns Change the Pattern use radio button to Pattern always I ll do this for graphs later in the chapter without further ado The default line patterns are solid short dashes and long dashes Click on line 3 in the right side of the dialog to select the 20 year treasury rate Then select the Line pattern dropdown menu and change the pattern to the very long dashes appearing at the end of the list EViews Illustrated book Page 123 Monday February 25 2013 10 06 AM 124 Chapter 5 Picture This Click to see the third ver sion of the graph which is virtu ally identical to the one that opened the chapter Graphic Auto Tweaks In making an aesthetically pleasing data graph EViews hides the details of many complex calculations Graphic output is tuned with many small tweaks to make the graph look just right In particular well done graphics scale nonlinearly In other words if you double the picture size picture elements don t simply grow to twice their original size If you plan to print or publish a graphic try to make it as close as possible to its final appearance while it s still inside EViews Pay particular attention to color versus monochrome and to the ratio of height versus width As an example we switched the graph above to be tall and
87. isn t in an EViews workfile one alternative is to use File Import which is designed to bring data into an existing workfile File Import is help Hint To prevent scrambling up the variables be sure to move data for all your series at the same time Hint You can use the right mouse menu item Insert obs to move the data for you automatically Hint In order to bring up the context menu with Paste as new Workfile be sure to right click in the lower EViews pane in a blank area i e outside of any workfile win dow EViews Illustrated book Page 55 Monday February 25 2013 10 06 AM 56 Chapter 2 EViews Meet Data ful if the data are in a spreadsheet or a text file It uses dialogs similar to the dialogs for Spreadsheet Read or ASCII Read from File Open Foreign Data as Workfile One more clipboard alternative makes you do more work but gives you a good deal of man ual control Use the Quick menu at the top of the EViews menu bar for the command Quick Edit Group Empty Series As you might infer from the suggestive name this brings up an empty group window all ready for you to enter data for one or more series You can paste anywhere in the window Essentially you are working in a spreadsheet view giving you complete manual control over editing This is handy when you only have part of a series or when you re gluing together data from different sources Among the Missing Mostly data are numbers Sometimes data are str
88. lation A Durbin Watson close to 2 0 is consistent with no serial correlation while a number closer to 0 means there probably is serial correlation The DW as the statistic is known of 0 095 in this example is a very strong indicator of serial correlation EViews has extensive facilities both for testing for the presence of serial correlation and for correcting regressions when serial correlation exists We ll look at the Durbin Watson as well as other tests for serial correlation and correction methods later in the book See Chapter 13 Serial Correlation Friend or Foe A Multiple Regression Is Simple Too Traditionally when teaching about regression the simple regression is introduced first and then multiple regression is presented as a more advanced and more complicated tech nique A simple regression uses an intercept and one explanatory variable on the right to explain the dependent variable A multiple regression uses one or more explanatory vari ables So a simple regression is just a special case of a multiple regression In learning about a simple regression in this chapter you ve learned all there is to know about multiple regres sion too Well almost The main addition with a multiple regression is that there are added right hand side variables and therefore added rows of coefficients standard errors etc The model we ve used so far explains the log of NYSE volume as a linear function of time Let
89. lower case letters are not equal The next comparison uses the function upper which produces an uppercase version of a string to pick up observations 1 through 5 by making the compari son using data converted to uppercase Both upper and lower case are changed to upper case before the comparison is made Converting everything to upper or lower case before making comparisons is a useful trick but doesn t help with fundamentally different spellings such as Wash and Washington For this particular set of data all spellings of Washington begin with wa and the spelling of no other state begins with wa So various spellings can be picked out by looking only at the first two letters in uppercase which is what the function left state 2 does for us left a n picks out the first n letters of string a EViews provides a comprehensive set of such functions As we said above see Appendix F of the Command and Programming Reference Embarrassing revelation I once spent weeks on a consulting project producing wrong answers because I didn t realize Washington had been spelled about half a dozen different ways The first few hundred observations that I checked visually all had the same spelling and I didn t think to check further EViews Illustrated book Page 104 Monday February 25 2013 10 06 AM Numbers and Letters 105 There s no general procedure for comparing strings by meaning rather than spelling al
90. mean a b b b ut R2 R2 R2 R2 1 R2 0 R2 EViews Illustrated book Page 66 Monday February 25 2013 10 06 AM The Pretty Important But Not So Important As the Last Section s Regression Results 67 t Tests and Stuff All the stuff about individual coefficients is reported in the middle panel a copy of which we ve yanked out to examine on its own The column headed t Statistic reports not surprisingly the t statistic Specifically this is the t statistic for the hypothesis that the coefficient in the same row equals zero It s com puted as the ratio of the estimated coefficient to its standard error e g Given that there are many potentially interesting hypotheses why does EViews devote an entire column to testing that specific coefficients equal zero The hypothesis that a coeffi cient equals zero is special because if the coefficient does equal zero then the attached coef ficient drops out of the equation In other words is really the same as with the time trend not mattering at all If the t statistic reported in column four is larger than the critical value you choose for the test the estimated coefficient is said to be statistically significant The critical value you pick depends primarily on the risk you re willing to take of mistakenly rejecting the null hypothesis the technical term is the size of the test and secondarily on the degrees of freedom for
91. more than pushing the button EViews Illustrated book Page 333 Monday February 25 2013 10 06 AM 334 Chapter 13 Serial Correlation Friend or Foe EViews Illustrated book Page 334 Monday February 25 2013 10 06 AM Chapter 14 A Taste of Advanced Estimation Estimation is econometric software s raison d tre This chapter presents a quick taste of some of the many techniques built into EViews We re not going to explore all the nuanced variations If you find an interesting flavor visit the User s Guide for in depth discussion Weighted Least Squares Ordinary least squares attaches equal weight to each observation Sometimes you want cer tain observations to count more than others One reason for weighting is to make sub popu lation proportions in your sample mimic sub population proportions in the overall population Another reason for weighting is to downweight high error variance observa tions The version of least squares that attaches weights to each observation is conveniently named weighted least squares or WLS In Chapter 8 Forecasting we looked at the growth of currency in the hands of the public estimating the equation shown here We used ordinary least squares for an esti mation technique but you may remember that the residuals were much noisier early in the sample than they were later on We might get a better estimate by giving less weight to the early observations EViews Illustrated book
92. not valid when the regressors include lagged dependent variables and it cannot be easily generalized to test for higher order processes While we recommend the more modern Breusch Godfrey in place of the Durbin Watson the truth is that the tests usually agree Breusch Godfrey Statistic The preferred test statistic for checking for serial correlation is the Breusch Godfrey From the menu choose Residual Diagnostics Serial Correlation LM Test to pop open a small dialog where you enter the degree of serial correlation you re interested in testing In other words if you re interested in first order serial correlation change Lags to include to 1 Hint EViews doesn t compute p values for the Durbin Watson Econometric warning But never use the Durbin Watson when there s a lagged depen dent variable on the right hand side of the equation DW 2 2r r 1 DW 2 EViews Illustrated book Page 320 Monday February 25 2013 10 06 AM Testing for Serial Correlation 321 The view to the right shows the results of testing for first order serial correla tion The top part of the output gives the test results in two versions an F statistic and a statistic There s no great reason to prefer one over the other Associated p values are shown next to each statistic For our stock market volume data the hypothesis of no serial correlation is easily rejected The bottom part of the view provides
93. not with a quarter number What s more the data runs backwards with the most recent observation com ing first Drop and drag this file onto EViews and EViews will not only figure out that it s quarterly data it ll also re sort the data into the right order so that it looks like the data shown to the right While EViews won t always figure out the intent of dates in a file it gets it right quite frequently so it s worth a try By the way this works with many for mats of input files not just Excel Reading the Great Texts If the second lowest common denominator file for mat is a Microsoft Excel spreadsheet what s the low est common denominator file format A plain text file of course Tab delimited Here s an excerpt from academic salaries by discipline txt available at the usual website Hint Text data and files are variously described as text ASCII alpha alpha numeric or character Sometimes the file extension csv is used for text data where data values are separated by commas EViews Illustrated book Page 43 Monday February 25 2013 10 06 AM 44 Chapter 2 EViews Meet Data As displayed the symbol represents a tab character the raised dot denotes a space and the paragraph mark marks the end of a line EViews interprets the tab character as separating one datum from the next and displays the text lined u
94. of the error from the previous period persists into the current period In contrast if nothing persists and there isn t any serial correlation If left untreated serial correlation can do two bad things Reported standard errors and t statistics can be quite far off Under certain circumstances the estimated regression coefficients can be quite badly biased When treated three good things are possible Standard errors and t statistics can be fixed The statistical efficiency of least squares can be improved upon Much better forecasting is possible We begin this chapter by looking at residuals as a way of spotting visual patterns in the regression errors Then we ll look at some formal testing procedures Having discussed detection of serial correlation we ll turn to methods for correcting regressions to account for serial correlation Lastly we talk about forecasting Fitted Values and Residuals A regression equation expresses the dependent variable as the sum of a modeled part and an error Once we ve computed the estimated regression coefficient we can make an analogous split of the dependent variable into the part explained by the regres sion and the part that remains unexplained The explained part is called the fitted value The unexplained part is called the residual The residuals yt bxt ut ut rut 1 e t 0 r 1 lt ut rut
95. only by context 2 Data where the format differs from one observation to another 3 Dates that aren t in English Verbum ianuarius non intellego Either read in the data the best you can and make corrections later or re arrange the data before you read it in Adding Data To An Existing Workfile Or Being Rectangular Doesn t Mean Being Inflexible You have a workfile set up and you ve populated it with data How you may ask do I add more data It helps to split this into two separate questions How do I add more observations and how do I add more variables In thinking about these questions picture your data as being formed into a rectangle and then lengthening the rectangle from top to bottom adding more observations or widening the rectangle from left to right adding more variables Pretend that our academic salary initially had only two variables DISCIPLINE and SALARY and five observations workfile range 1 5 We could picture the data as being in a rectan gle with two columns and five rows We now discover a sixth observation Geology with a salary of 33 206 Putting in the new observation takes two steps First we extend the workfile at the bottom Next we enter the new observation in the space we ve created Hint If your data are truly irregular it s possible that reading it directly into EViews is just not a happening event You may be better off lightly touching up the data organi zation in a tex
96. regression results suggest that Asians earn two and a half percent more than the rest of the population after accounting for age and education although the significance level is a smidgen short of the 5 percent gold standard However the Asian population isn t distrib uted randomly across the United States If Asians are relatively more likely to live in high wage areas our regression might be unintentionally picking up a location effect rather than a race effect We can look at this issue by including dummy variables for the 51 states DC too eh This can be done directly and we ll do it directly in the next section but it can be very conve nient to pretend that each state identifies a cross section in a panel so that we can use EViews panel estimation tools In other words let s fake a panel Double click on Range to bring up the Workfile structure dia log Set the dialog to Undated Panel since there aren t any dates and uncheck Balance between starts and ends since there isn t anything to balance Now that we have a pretend panel we can re run the estima tion and then use the estimation options to include cross section fixed effects Junk science alert Asian Americans are an especially diverse socio economic group Other than the American tendency to use race to classify everything it s not clear why fourth and fifth generation Japanese Americans should be lumped together with recent Hmong refugees Fo
97. s add two more variables time squared and lagged log volume hoping that time and time squared will improve our ability to match the long run trend and that lagged values of the dependent variable will help out with the short run In the last example we entered the specification in the Equation Estimation dialog I find it much easier to type the regression command directly into the command pane although the Parallel construction notice The fourth and fifth columns in EViews regression output report the t statistic and corresponding p value for the hypothesis that the individual coefficient in the row equals zero The F statistic in the summary area is doing exactly the same test for all the coefficients except the intercept together This example has only one such coefficient so the t statistic and the F statistic test exactly the same hypothesis Not coincidentally the reported p values are identical and the F is exactly the square of the t 2672 51 72 EViews Illustrated book Page 73 Monday February 25 2013 10 06 AM 74 Chapter 3 Getting the Most from Least Squares method you use is strictly a matter of taste The regression command is ls followed by the dependent variable followed by a list of independent variables using the special symbol C to signal EViews to include an intercept In this case type ls log volume c trend trend 2 log volume 1 and EViews brings up the multiple regression output sh
98. sam ple from the x axis Notice that it looks like the dis tance between 2 and 4 is the same as the distance between 1 and 2 or 4 and 5 This makes sense if the x coordinates are ordinal but isn t so good if they re cardi nal In other words drop ping part of the x axis works if the measurements are things like strongly agree agree indiffer ent but doesn t work so well for measurements like 1 mile from the Eiffel Tower 2 miles from the Eiffel Tower etc Pad excluded obs leaves in the part of the x axis for which data are missing It s better for cardinal ordinates EViews Illustrated book Page 179 Monday February 25 2013 10 06 AM 180 Chapter 6 Intimacy With Graphic Objects Segment with lines is a stronger version of Drop excluded deleting more of the missing x axis with an added vertical lines to show the break points Segment with lines is the most intellectually honest dis play because it makes sure everyone knows where breaks in the sample occur The two disadvantages are that it s not always the most aesthetically pleasing pic ture especially if there are lots of breaks and that the vertical line draws attention to the sample break which may or may not be a particularly interesting part of the data Hint Checking Connect adjacent for either of the first two choices connects points on either side of the sample break This fre quently make
99. share a maximum length Strings with more than the maximum permitted charac ters are truncated By default 40 char acters are permit ted While you can t change the trunca tion limit for an indi vidual series you can change the default through the General Options Series and Alphas Alpha truncation dialog And you should Again see Chapter 18 Optional Ending Hint If you use text in the command line be sure to enclose the text in quotes For example alpha alphabet abc s To include a quote symbol in a string as part of a command use a pair of quotes To create a string consisting of a single double quote symbol write alpha singledoublequote Hint Named auto series defined using frml work for alpha series just as they do for numeric series EViews Illustrated book Page 103 Monday February 25 2013 10 06 AM 104 Chapter 4 Data The Transformational Experience String Functions String functions are documented in the Command and Program ming Reference An example here gives a taste The series STATE contains state names Unfortunately spellings vary Consider the following expressions show state state Wash upper state WASH upper left state 2 WA The group win dow shows the results of all three compari sons state Wash is true for observations 1 2 and 4 Wash but not for the third or fifth observation wash and WASH because upper and
100. sion model and use it for forecasting Workfile The Basic EViews Document Start up a word processor and you re handed a blank page to type on Start up a spreadsheet program and a grid of empty rows and col umns is provided Most pro grams hand you a blank document of one sort or another When you fire up EViews you get a welcome screen offering you some choices about how you d like to get started To get some support before diving into EViews you can turn your attention to the section on the right which offers various tutorials and online help Being impatient to get started let s take the quick solution and load an existing workfile If you re working on the computer while reading you may want to load the workfile nysevol ume wf1 by clicking on the Open an existing EViews workfile button If you have saved the workfile on your computer navigate to its location and open it While word processor documents can start life as a generic blank page EViews docu ments called workfiles include information about the structure of your data and there fore are never generic Consequently creating an EViews workfile and entering data takes a Hint All the files used in this book are available on the web at www eviews com EViews Illustrated book Page 3 Monday February 25 2013 10 06 AM 4 Chapter 1 A Quick Walk Through couple of minutes or at least a couple of seconds of explanation In the nex
101. skinny See Frame amp Size in the next chap ter Notice that EViews switched the tick labeling on both horizontal and vertical axes to keep the labeling looking pretty The aesthetic choices made by EViews involve complicated interac tions between space font sizes and other factors Although all the options can be controlled manually it s generally best to trust EViews judgment Print Copy Or Export It s not hard to print a copy of a graph just click the button If you re sending to a color printer check the Print in Color checkbox in the print dialog EViews Illustrated book Page 124 Monday February 25 2013 10 06 AM A Simple Soup To Nuts Graphing Example 125 To make a copy of a graph object inside an EViews workfile use Object Copy Object or copy and paste the object in the workfile window To make an external copy you can either copy and paste or save the graph as a file on disk Graph Copy and Paste With the graph window active you copy the graph onto the clipboard by hitting the button and choosing Copy or by choosing Copy from the right click menu or with the usual Ctrl C Then paste into your favorite graphics program or word processor EViews pic tures are editable so a certain amount of touch up can be done in the destination program If you d like to do even more editing after your graph is in your source document you should paste as an embedded graph Embedded graphs are im
102. something that s already happened Practical hint If you re like the author you usually set up the range of your EViews workfile to coincide with your data sample Then when I forecast out of sample I get an error message saying the forecast sample is out of range When this happens to you just double click Range in the upper panel of the workfile window to extend the range to the end of your forecasting period TL T TF T gt x EViews Illustrated book Page 227 Monday February 25 2013 10 06 AM 228 Chapter 8 Forecasting Dynamic Versus Static Forecasting Our currency data ends in April 2005 To forecast past that date we need to know the value for TREND no problem which month we re forecasting for no problem and the value of currency growth in the month previous to the forecast month maybe a problem It s this lagged dependent variable that presents a problem opportunity If we re forecasting for May 2005 we re okay because we know the April value But for June or later we don t have the lagged value of currency growth A static forecast uses the actual values of the explanatory variables in making the forecast In our example we can make static forecasts through May 2005 but no later A dynamic forecast uses the forecast value of lagged dependent variables in place of the actual value of the lagged dependent variables If we start a forecast in May 2005 the dynamic forecast is identical to
103. t 51 70045 ser 0 967362 R2 0 852357 EViews Illustrated book Page 69 Monday February 25 2013 10 06 AM 70 Chapter 3 Getting the Most from Least Squares What s the use of the top three lines It s nice to know the date and time but EViews is rather ungainly to use as a wristwatch More seriously the top three lines are there so that when you look at the output you can remember what you were doing Dependent Variable just reminds you what the regression was explaining LOG VOLUME in this case Method reminds us which statistical procedure produced the output EViews has dozens of statistical procedures built in The default procedure for estimating the parameters of an equation is least squares Big Digression Hint Automatic exclusion of NA observations can sometimes have sur prising side effects We ll use the data abstract at the right as an example Data are missing from observation 2 for X1 and from observation 3 for X2 A regres sion of Y on X1 would use observations 1 3 4 and 5 A regression of Y on X2 would use observations 1 2 4 and 5 A regression of Y on both X1 and X2 would use obser vations 1 4 and 5 Notice that the fifth observation on Y is zero which is perfectly valid but that the fifth observation on log Y is NA Since the logarithm of zero is undefined EViews inserts NA whenever it s asked to take the log of zero A regression of log Y
104. the static forecast Both use as an explana tory variable The dynamic forecast for June uses The static forecast can t be com puted because isn t known The Role of the Forecast Sample In practice the difference between static and dynamic forecasting depends on both data availability and the specification of the forecast sample Static forecasting uses values of explanatory variables from the forecast sample If any of them are missing for a particular date then nothing gets forecast for that date Dynamic forecasting pretends that you don t have any information about the dependent variable during the period covered by the sample forecast even when you do have the rel evant data In the first period of the forecast sample EViews uses the actual lagged depen dent variables since these actual values are known In the second period EViews pretends it doesn t know the value of the lagged dependent variable and uses the value that it had just forecast for the first period In the third period EViews uses the value forecast for the sec ond period And so on One of the nice things about dynamic forecasting is that the fore Hint The makes all the difference here We always have because that s the number we forecast in period We re just rolling the forecasts forward In con trast once we re more than one period past the end of our data sample is unknown Hint The data used for the explanatory vari
105. the test The larger the risk you re willing to take the smaller the critical value and the more likely you are to find the coefficient significant The textbook approach to hypothesis testing proceeds thusly 1 Pick a size the probability of mistakenly rejecting say five percent 2 Look up the critical value in a t table for the specified size and degrees of freedom Foreshadowing hint EViews automatically computes the test statistic against the hypothesis that a coefficient equals zero We ll get to testing other coefficients in a minute but if you want to leap ahead look at the equation window menu View Coef ficient Tests Hint EViews doesn t compute the degrees of freedom for you That s probably because the computation is so easy it s not worth using scarce screen real estate Degrees of freedom equals the number of observations reported in the top panel on the output screen less the number of parameters estimated the number of rows in the middle panel In our example 51 7 0 017 0 00033 volumet log a 0 t ut volumet log a ut df 465 2 463 EViews Illustrated book Page 67 Monday February 25 2013 10 06 AM 68 Chapter 3 Getting the Most from Least Squares 3 Compare the critical value to the t statistic reported in column four Find the variable to be significant if the t statistic is greater than the critical value EV
106. to and so forth Adding up the first differences is the source of the term integrated Hint The first difference of LOGVOL can be written in two ways D LOGVOL or LOGVOL LOGVOL 1 The two are equivalent dy1 y1 y0 dy2 y2 y1 y1 dy1 y0 y2 dy2 dy1 y0 EViews Illustrated book Page 330 Monday February 25 2013 10 06 AM ARMA and ARIMA Models 331 ARIMA Estimation Here s a plot of D LOGVOL No more exploding Building an ARIMA model of LOGVOL boils down to building an ARMA model of D LOGVOL It s traditional to treat the depen dent variable in an ARMA model as having mean zero One way to do this is to use the expres sion D LOGVOL MEAN D LOGV OL as the dependent variable but it s just as easy to include a constant in the estimate So with that build up here s how to estimate an ARIMA 1 1 1 model in EViews ls d logvol c ar 1 ma 1 Hint Once in a great while it s necessary to difference the first difference in order to get a stationary time series If it s necessary to difference the data d times to achieve stationarity then the original series is said to be integrated of order d or to be an series The complete specification of the order of an ARIMA model is ARIMA p d q where a plain old ARMA model is the special case ARIMA p 0 q By the way the d function generalizes in EViews so that d y d is the dth differ ence of Y I d
107. to capture what you see you might want to consider freezing the graph To Freeze Or Not To Freeze Before adding any customizations we have a choice to make about whether to freeze the graph The group we re looking at right now is fundamentally a list of data series which we happen to be looking at in a graphics view If we change any of the underlying data the change will be reflected in the picture Same thing if we change the sample For that matter we can switch to a different graphics view or even a spreadsheet or statistical view But a group view does have one shortcoming Hint Where did those nice long descriptive names in the legend come from EViews automatically uses the DisplayName for each series in the legend if the series has one See Label View on page 30 in Chapter 2 EViews Meet Data If there is no Dis playName the name of the series is used instead Hint If you hover your cursor over a data point on a line in the graph EViews will show you the observation label and value If you hover over any other point inside the graph frame EViews will display the values in the statusline located in the lower left hand corner of your EViews window EViews Illustrated book Page 119 Monday February 25 2013 10 06 AM 120 Chapter 5 Picture This When you close a group window or shift to another view many customizations you ve made on the graph disappear Freezing a graph view creates a new obj
108. to use compression so use it unless someone using a ver sion of EViews earlier than 5 0 needs to read the file in which case don t EViews Illustrated book Page 411 Monday February 25 2013 10 06 AM 412 Chapter 18 Optional Ending File Locations As a rule EViews users never mess with the File locations settings But you may be excep tional since you re reading a chapter on setting options Power users sometimes keep around several differ ent sets of options each fine tuned for a particu lar purpose The EViews Paths dialog lets you pick a path for each option setting EViews automatically names its options storing file EViews32 ini To store multiple ver sions of EViews32 ini fine tune your options to suit a particular purpose then reset the Ini File Path to a unique path for each version you wish to save There s no browsing for EViews32 ini and the name EViews32 ini is hard coded into the program so to use multiple option sets you need to remember the paths in which you ve stored each set Graphics Defaults Going back to the Options menu and clicking on Graph ics Defaults you ll see the Graph Options dialog is enormous with many sections These options set the defaults used for options when you first create a graph The same tabs appear on the Graph Options dia log for an individ EViews Illustrated book Page 412 Monday February 25 2013 10 06 AM Quick Re
109. very small demonstrates the maxim that with enough data you can accurately identify differences too small for anyone to care about Statistical hint The basic Anova F test for differences of means assumes that the sub populations have equal variances Most introductory statistics classes also teach the Satterthwaite and Welch tests that allow for different variances for different subpopu lations EViews Illustrated book Page 213 Monday February 25 2013 10 06 AM 214 Chapter 7 Look At Your Data Describing Groups Just the Facts Putting It Together Many of the descriptive features for groups are the same as those for series except done for each of the series in the group For example the descriptive statistics menu for a group offers two choices Common Sample Individual Samples The choices produce basic descriptive sta tistics for each series with two different arrangements for choosing the sample If you want the same set of observations to go into computing the statistics for each series in the group be sure to specify Common Sample Using Individual Samples we can see that there were quite a few applicants with valid LSAT scores but not valid GPAs 1707 versus 1638 observations As a guess this could reflect applications from undergraduate schools that don t compute grade point averages The Tests of Equality view for groups checks whether the mean or median or variance is the same for all the series in t
110. we ll link the models to the equations you ve already learned to estimate Your First Homework Bam Taken Up A Notch Odds are that your very first homework assignment in your very first introductory macro economics class presented a model something like this Your assignment was to solve for the variables Y GDP and C consumption given infor mation about I investment and G government spending But that consumption function is econometrically pretty unsophisticated We ll carefully avoid any questions about the sophistication of a model consisting solely of a national income identity and a consumption function A more modern consumption function might look like this Cultural Imperialism Apologia If you took the course outside the United States the national income identity probably included net exports as well Y C I G C C mpc Y Ct ln a l Ct 1 b Yt ln ln EViews Illustrated book Page 365 Monday February 25 2013 10 06 AM 366 Chapter 15 Super Models The page Real in the workfile Key nes wf1 contains annual national income accounting data for the U S from 1959 through 2000 It also includes an estimated equation for this more modern consumption function Creating A Model This model isn t so easy to solve as is the Keynesian cross The consump tion function introduces both nonlin ear and dynamic elements Fortunately this sort of
111. where union bars are shown in solid red and nonunion bars come in cross hatched blue Turning on Within graph category identification instructs EViews to add visual distinction to the within graph elements We ve done this to change Hint Categorical options aren t available on frozen graphs if updating is disabled EViews Illustrated book Page 162 Monday February 25 2013 10 06 AM Categorical Graphs 163 the color in the graph to the right by changing Within graph category identification from none to UNION EViews interprets add visual distinction for this graph as assigning a unique color for each within category This choice is ideal when the graph is presented in color but we wanted clear visual distinction for a monochrome version as well So we added the cross hatching manually To see how see Fill Areas in the next chapter More Polish Two more items are worthy of quick mention The first item is that you can direct EViews to maximize the use of either axis labels or legends Axis labels are generally better than using a legend but sometimes they just take up too much room The graph above maximized label use Here s the same graph maximizing legend use In this example neither graph is hard to understand nor terribly crowded so which one is better mostly depends on your taste EViews Illustrated book Page 163 Monday February 25 2013 10 06 AM 164 Chapter 5 Picture This The second item worth k
112. which itself has well over a dozen sections If you prefer you can reach the same features by right clicking anywhere in the graph window and choosing from the context menu The same menu along with a few others features pops up if you click on the button The menu provides an editing feature that isn t replicated with a button Draw arrow Hint Double click on almost anything in a graphics window and the appropriate dia log will pop open presenting you with the options available for fine tuning Hint Double click on nothing in a graphics window i e in a blank spot and the Graph Options dialog will pop open This is probably the easiest way to reach all the fine tuning options EViews Illustrated book Page 167 Monday February 25 2013 10 06 AM 168 Chapter 6 Intimacy With Graphic Objects To Freeze Or Not To Freeze Redux You ll remember from the last chapter that a graph in a series or group window can be fro zen to create a graph window Freezing can be done with updating turned on or updating turned off freezing with updating turned off severs the graph from the original data In general it s better to freeze a graph before fine tuning One option discussed in the cur rent chapter sorting is only available after freezing with updating turned off On the other hand some functions that you might think of as fine tuning EViews regards as part of the graph creation process Adding axis borders is one example T
113. workfile window and you re in business Multiple Frequencies Multiple Pages In Chapter 2 EViews Meet Data we wrote It is now later in the book All the data in a page within a workfile share a common identifier For example one page might hold quarterly data and another might hold annual data So if you want to keep both quarterly and annual data in one workfile set up two pages Even better you can easily copy data from one page to another converting frequencies as needed The workfile US Output wf1 available on the EViews website holds data on the U S Index of Industrial Production IP and on real Gross Domestic Product RGDP In contrast to GDP numbers which are computed quarterly industrial production numbers are avail able monthly For this reason industrial production is the output measure of choice for com parison with monthly series such as unemployment or inflation Since the industrial Hint Every variable in an EViews workfile shares a common identifier series You can t have one variable that s measured in January February and March and a differ ent variable that s measured in the chocolate mixing bowl the vanilla mixing bowl and the mocha mixing bowl Subhint Well yes actually you can EViews has quite sophisticated capabilities for handling both mixed frequency data and panel data These are covered later in the book Hint EViews will convert between frequencies automat
114. 013 10 06 AM 142 Chapter 5 Picture This The Overlap option this will not come as a great surprise allows the lines to overlap Since the lines share the vertical space they re each a little easier to read The downside is that the viewer s attention may be drawn to the line crossings which for these series aren t meaningful Now let s take a look at some graph types that apply only when there s more than one series Area Band Area Band plots a band using pairs of series by filling in the area between the two sets of val ues Band graphs are most often used to display forecast bands or error bands around forecasts EViews will construct bands from successive pairs of series in the group If there is an odd number of series in the group the final series will by default be plotted as a line The example here shows high and low prices for the Russell 3000 a very broad index of U S stocks in the later part of 1987 EViews Illustrated book Page 142 Monday February 25 2013 10 06 AM Group Graphics 143 Mixed With Lines Mixed with Lines plots the first series in a group as an Area Bar or Spike graph your choice and the remaining series as lines We ve used this feature to put the U S national debt and GDP together in one picture Note that we ve used both left and right axes for labels Note too how nicely mixing an Area and a Line graph illustrates that debt dropped dramat
115. 346 L label views 30 lags 87 89 280 LaTeX 125 413 least squares estimation 306 335 338 341 347 359 360 regression 275 276 See also ls command left function 104 left hand side variable 65 left to right order changing 32 legends 119 187 188 letter data type 102 103 limited dependent variables 349 351 line graphs 6 131 140 142 See also seasonal graphs linear match last conversion 249 lines in graphs 122 123 171 172 173 175 188 192 links EViews Illustrated book Page 419 Monday February 25 2013 10 06 AM 420 Index contracting 261 264 creating 252 253 defined 251 252 matching with 256 261 NA values and 267 series names and 267 timing of 267 VARs and 379 Ljung Box Q statistic 321 322 locking unlocking windows 29 logarithmic scales 148 logarithms 90 logical operators 86 87 logit model 350 351 loops program 385 386 387 low to high frequency conversion 247 248 ls command 14 17 74 M MA moving average errors 322 323 324 325 See also autoregressive moving average ARMA errors makedate function 107 108 matching 255 256 See also links mathematical functions 90 maximum likelihood estimation mle 355 357 mean graphing 129 mean function 99 112 mean merges 261 263 mean testing 211 means removed statistics with 299 300 meansby function 112 113 266 measurements data 100 missing 34 median testing 211 mediansby function 112
116. 6 Chapter 8 Forecasting In other words Step 3 consists of multiplying each estimated coefficient from Step 2 by the relevant value in the forecast period and then adding up the products The formalism will turn out to be convenient for our discussion below For now let s use the example we ve been working with to forecast for November 2004 The variables in our equation are a time trend growth the previous month and a set of dummy 0 1 variables for the month To forecast we need the values of each variable and the estimated coefficient attached to that variable The required data appear in the fol lowing table When you click the button EViews does the same calculation just faster and with some extra doo dads available in the output In Sample and Out Of Sample Forecasts In carrying out Step 3 there s an implicit issue that deserves explicit attention Before we could start multiplying parameters times variables in the table above we needed to know the values of in the forecast period In our table the value of TREND is 1047 the month number in our data for November 2004 We cheated and looked the number up by giving the command show trend which opens a spreadsheet view of TREND The November value of lagged currency growth is the October currency growth number 3 4105 And MONTH 11 always equals 1 0 in November Suppose we had estimated a better model that used the inflation rate as an explanatory
117. 6 AM Saving your work 19 Before leaving the computer click on the workfile window Use the File menu choice File Save As to save the workfile on the disk Now would be a good time to take a break In fact take a few minutes and indulge in your favorite beverage Back so soon If your computer froze while you were gone you can start up EViews and use File Open EViews Workfile to reload the workfile you saved before the break Your computer didn t freeze did it But then you probably didn t really take a break either This is the spot in which authors enjoin you to save your work often The truth is EViews is remarkably stable software It certainly crashes less often than your typical word processor So yes you should save your workfile to disk as a safety measure since it s easy but there s a different reason that we re emphasizing saving your workfile EViews doesn t have an Undo feature As you work you make changes to the data in the workfile Sometimes you find you ve gone up a blind alley and would like to back out Since there is no Undo feature we have to sub stitute by doing Save As frequently If you like save files as foo1 wf1 foo2 wf1 etc If you find you ve made changes to the workfile in memory that you now regret you can backup by loading in foo1 wf1 You can also hit the Save button on the workfile window to save a copy of the workfile to disk This is a few ke
118. CAN and GDPUSA In a panel in contrast the structure of the data applies to all series in the workfile One way to think about the difference between the two structures is seen in the steps needed to include a particular cross section in an analysis For a panel all cross sections are included except for ones you exclude through a smpl statement In a pool only those cross sections identified in the pool are included and a smpl statement is used only for the time dimension The flip side of this is that a panel has one fundamental structure built into the workfile while in the pool setup you can define as many different pool objects as you like Nuances If you re thinking that pools are easier to learn about you re absolutely right But panels provide more powerful tools Here s another rule of thumb If you think of yourself as a time series person you ll probably find pools the more natural concept but if you re a cross section type then try out a panel first Here s a practical rule Historical hint Pools have been part of EViews for a long time panels are a relatively new feature Hint Pools are designed for handling a modest number of time series bundled together while panels are better for repeated observations on large cross sections One rule of thumb is that data in which an individual time series has an interesting iden tity Canada for example is likely to be a candidate to be trea
119. Creation The methods above begin with an existing series or a whole handful of existing series and create links from them You can also create a link in the active page by choosing the Object menu or the but ton and selecting New Object Series Link Or you can type link in the command pane optionally fol lowed by a name for the link If you specify a name a new link appears in the workfile window although the link isn t yet actually linked to anything If you don t specify a name EViews pops open a view win dow for the new link and displays an annoying error message Unable to Perform Link Ignore the message and switch to the Properties view Click on the button switch to the Link Spec tab and enter the source series and source workfile page in the Link to field You can also specify the frequency method conversion to be used in the Frequency conversion options field Auto Link Or you might not need to create a link object in your workfile page at all Anywhere that you can use a series you can also use a series from another workfile page If you are creat Hint If you ve copied multiple series onto the clipboard you can use the and buttons in the Paste Special dialog to paste the series in one at a time or to paste them in all at once EViews Illustrated book Page 253 Monday February 25 2013 10 06 AM 254 Chapter 9 Page After Page After Page ing a link for use in an equati
120. ELATION FRIEND OR FOE 315 Visual Checks 317 Testing for Serial Correlation 319 More General Patterns of Serial Correlation 322 Correcting for Serial Correlation 323 Forecasting 327 ARMA and ARIMA Models 329 Quick Review 333 CHAPTER 14 A TASTE OF ADVANCED ESTIMATION 335 Weighted Least Squares 335 Heteroskedasticity 338 Nonlinear Least Squares
121. EViews Illustrated for Version 8 Richard Startz University of California Santa Barbara EViews Illustrated book Page 1 Monday February 25 2013 10 06 AM EViews Illustrated for Version 8 Copyright 1994 2013 IHS Global Inc All Rights Reserved ISBN 978 1 880411 19 3 This software product including program code and manual is copyrighted and all rights are reserved by IHS Global Inc The distribution and sale of this product are intended for the use of the original purchaser only Except as permitted under the United States Copyright Act of 1976 no part of this product may be reproduced or distributed in any form or by any means or stored in a database or retrieval system without the prior written permission of IHS Global Inc Disclaimer The authors and IHS Global Inc assume no responsibility for any errors that may appear in this manual or the EViews program The user assumes all responsibility for the selection of the pro gram to achieve intended results and for the installation use and results obtained from the pro gram Trademarks EViews is a registered trademark of IHS Global Inc Windows and Excel are registered trade marks of Microsoft Corporation All other product names mentioned in this manual may be trademarks or registered trademarks of their respective companies IHS Global Inc 4521 Campus Drive 336 Irvine CA 92612 2621 Telephone 949 856 3368 Fax 949 856 2044 e mail sales eviews com
122. Frequently the best thing to do with data you don t have is nothing at all EViews statistical procedures offer a variety of options but the usual default is to omit NA observations from the analysis EViews Illustrated book Page 56 Monday February 25 2013 10 06 AM Quick Review 57 by filling out the field Text representing NA in the dialog used to read in the data EViews allows only one value to be automatically translated this way Reading alpha series with missing values is slightly more problematic because any string of characters might be a legitimate value Maybe the characters NA are an abbreviation for North America For an alpha series you must explicitly specify the string used to represent missing data in the Text representing NA field Handling multiple missing codes Some statistical programs allow multiple values to be considered missing Others EViews being a singular example permit only one code for missing values Suppose that for some variable call it X the values 9 99 and 999 are all sup pose to represent missing data The way to han dle this in EViews is to read the data in without specifying any values as missing and then to recode the data In this example this could be done by choosing Quick Generate Series and then using the Generate Series by Equa tion dialog to set the sample to include just those values of x that you want recoded to NA If you prefer you can accomplish the sa
123. GDP with the con version Up Frequency Conversions Hold on one second If GDP is only measured quarterly how did we manage to use GDP data in a monthly regression After all monthly GDP data doesn t exist If you were hoping the good data fairy waived her monthly data wand and will do the same for you well sorry that s not what happened Instead EViews applied a low to high frequency data conversion rule In this case EViews used a default rule Let s go over what rules are available discuss how you can make your own choice rather than accept the default and then look at how the defaults get set RGDP 338 92 IP EViews Illustrated book Page 247 Monday February 25 2013 10 06 AM 248 Chapter 9 Page After Page After Page As background let s take a look at quarterly GDP values for the year 2004 and the monthly values manufactured by EViews We can see that EViews copied the value of GDP in a given quarter into each month of that quarter That s a reasonable thing to do If all we know was that GDP in the first quarter was running at a rate of a trillion dollars a year then saying that GDP was running at a trillion a year in January a trillion a year in February and a trillion a year in March seems like a good start What else could we have done The full set of options is shown at the right The first choice Specified in series isn t an actual conversion method It signals to use which
124. Hakuna matata Probably half the graphs ever produced in EViews are line graphs As you ve already seen this requires you to 1 Open a window with desired data 2 Choose Graph from the View menu 3 Click EViews Illustrated book Page 127 Monday February 25 2013 10 06 AM 128 Chapter 5 Picture This Or suppose we wanted a scatter plot of long rates against the short 3 month rate Just choose the Scatter and accept the defaults settings to display the scatterplots EViews Illustrated book Page 128 Monday February 25 2013 10 06 AM A Graphic Description of the Creative Process 129 But here s my favorite one click wonder Change the Graph type back to Line amp Symbol and then with a single click change Graph data to Means EViews Illustrated book Page 129 Monday February 25 2013 10 06 AM 130 Chapter 5 Picture This The graph now flicks from raw data to a particularly interesting summary Instead of a line graph for each series EViews plots the mean for each series and connects the means with a line This view of interest rates called an average yield curve shows at a glance that long term interest rates are typi cally higher than short term rates with 20 year bonds paying on aver age about 2 5 percentage points more than 3 month bonds My favorite two click wonder takes the previous graph and adds a click on Bar to give us this version of the same summary information Financial eco
125. Here s the Wald Test view after entering c 2 0 c 3 0 The hypothesis is rejected Note that EViews correctly reports 2 degrees of freedom for the test statistic Representing The Representations view shown at the right doesn t tell you anything you don t already know but it provides useful reminders of the com mand used to generate the regression the interpretation of the coefficient labels C 1 C 2 etc and the form of the equation written out with the estimated coefficients What s Left After You ve Gotten the Most Out of Least Squares Our regression equation does a pretty good job of explaining log volume but the explana tion isn t perfect What remains the difference between the left hand side variable and the value predicted by the right hand side is called the residual EViews provides several tools to examine and use the residuals Hint Okay okay Maybe you didn t really need the representations view as a reminder The real value of this view is that you can copy the equation from this view and then paste it into your word processor or into an EViews batch program or even into Excel where with a little judicious editing you can turn the equation into an Excel formula t t2 EViews Illustrated book Page 78 Monday February 25 2013 10 06 AM What s Left After You ve Gotten the Most Out of Least Squares 79 Peeking at the Residuals The View Actual Fitted Residual provid
126. MA 1 use the command Hint Unlike nearly all other EViews estimation procedures MA requires a continuous sample If your sample includes a break or NA data EViews will give an error mes sage e u R2 r1 r2 r1 EViews Illustrated book Page 324 Monday February 25 2013 10 06 AM Correcting for Serial Correlation 325 ls logvol c trend trend 2 d log close 1 ar 1 ar 2 ma 1 The results shown to the right give both autoregressive coefficients and the single moving average coefficient All three ARMA coefficients are signif icant Another Way to Look at the ARMA Coefficients Equations that include ARMA parameters have an ARMA Structure view which brings up a dialog offering four diagnostics We ll take a look at the Correlogram view here and the Impulse Response view in the next section EViews Illustrated book Page 325 Monday February 25 2013 10 06 AM 326 Chapter 13 Serial Correlation Friend or Foe Here s the correlogram for the volume equation estimated above with an AR 1 specification The correlogram shown in the top part of the figure uses a solid line to draw the theoretical correlogram corresponding to the esti mated ARMA parameters The spikes show the empirical correlogram of the residuals the same values as we saw in the residual correlogram earlier in the chapter The solid line theoretical and the top of the spikes empirical don t match up very well do
127. Models Our new graph which we have prettied up shows both baseline and scenario 1 results Putting the deviation on a separate scale makes it easier to see the effect of this fiscal policy experiment With a little luck these results will get us a very good grade Simulating VARs Models can be used for solving complicated systems of equations under different scenarios as we ve done above Models can also be used for forecasting dynamic systems of equa tions This is especially useful in forecasting from vector autoregressions Open the workfile currency model wf1 which contains the vector autoregression estimated in Chapter 14 A Taste of Advanced Estimation Create a new model object named CURRENCY_FORECAST Copy the VAR object CASH from the workfile window and paste it into the model which now looks as shown to the right Hint The feature in equations handles dynamic forecasting from a single equation quite handily Here we re talking about forecasting from multiple equation models EViews Illustrated book Page 378 Monday February 25 2013 10 06 AM Simulating VARs 379 Double click on the equa tion The equation for G growth in currency in the hands of the public has been copied in together with the estimated coefficients and the estimated standard error of the error terms in the equation Remember this is a live link so If you re estimate the VAR the model will know to use the re
128. Monday February 25 2013 10 06 AM Describing Series Just The Facts Please 201 Stats Table The menu Descriptive Statistics Stats Table creates a table with pretty much the same information as is found in the statis tics panel of Histogram and Statistics This table format has the advantage that it s easier to copy and paste into your word pro cessor or spreadsheet program Stats By Classification A common first step on the road from data description to data analysis is asking whether the basic series statistics differ for sub groups of the population Clicking Descriptive Statis tics Stats by Classification brings up the Statistics By Clas sification dialog You ll see a field called Series Group for classify smack in the upper cen ter of the dialog Enter one or more series or groups here hit and you get summary statistics computed for all the distinct combinations of values of the classifying series Here s a simple example In our workfile the vari able WASH equals one for Washington State resi dents and zero for everyone else Using WASH as the classifying variable gives the results shown to the right About 60 percent of applications 1028 out of 1639 were from out of the state and the out of state applicants averaged a slightly higher GPA EViews Illustrated book Page 201 Monday February 25 2013 10 06 AM 202 Chapter 7 Look At Your Data If we wanted to see the effect of state and having
129. Monday February 25 2013 10 06 AM Numbers and Letters 101 Number Display By default EViews displays numbers in a format that s pretty easy to read You can change the format for displaying a series through the Properties dialog of the spreadsheet view of the series Check boxes let you use a thousands separator commas are the default put negative values in parentheses often used for financial data add a trailing doesn t change the value displayed just adds the percent sign or use a comma to separate the integer and decimal part of the number The Numeric display dropdown menu in the dialog provides options in addition to the default Fixed characters Significant dig its drops off trailing zeros after the decimal Using Fixed decimal you can pick how many places to show after the decimal For exam ple you might choose two places after the decimal for data mea sured in dollars and cents Scientific notation sometimes called engineering notation puts one digit in front of the decimal and includes a power of ten For example 12 05 would appear as 1 205E01 Percentage displays numbers as percentages so 0 5 appears as 50 00 To add a symbol check the Trailing checkbox Fraction displays fractions rather than deci Hint Click on a cell in the spreadsheet view to see a number displayed in full pre cision in the status line at the bottom of the EViews window EViews Illustrated book Page 101 Mo
130. Nifty About Panel Data Panel data presents two big advantages over ordinary time series or cross section data The obvious advantage is that panel data frequently has lots and lots of observations The not always obvious advantage is that in certain circumstances panel data allows you to control for unobservables that would otherwise mess up your regression estimation Panels can be big It s helpful to think of the observations in a time series as being numbered from 1 to T even though EViews typically uses dates like 2004q4 rather than as identifiers Cross section data are numbered from 1 to N it being something of a convention to use T for time series and N for cross sections Using i to subscript the cross section and t to subscript the time period we can write the equation for a regression line as With a panel we are able to estimate the regression line using observations which can be a whole lot of data leading to highly precise estimates of the regression line For example the Penn World Table Alan Heston Robert Summers and Bettina Aten Penn World Table Version 6 1 Center for International Comparisons at the University of Pennsyl vania CICUP October 2002 has data on 208 countries for 51 years for a total of more than 10 000 observations We ll use data from the Penn World Table for our first examples Using panels to control for unobservables A key assumption in most applications of least squares regression i
131. Page 257 Monday February 25 2013 10 06 AM 258 Chapter 9 Page After Page After Page We can now use REV just like any other series EViews will bring data in from the Revenue page each time it s needed For example a scatter diagram of infant mortality against per capita revenue shows a slight and surprising positive association The positive association is attribut able to the one outlier Drop Alaska and the picture shifts to a slight neg ative relation In this example we ve used links to match in a case where there really was a common identifier the com puter just didn t know it Next we turn to matching up series with fundamentally different identifiers Matching When The Identifiers Are Really Different In this next example our main data set holds observations on individuals We re going to hook up these individual observations with data specific to each person s state of residence In order to show off more EViews features we ll generate the state by state data by taking averages from the individual level data For a real problem to work on we re going to try to answer whether higher unionization rates raise wages for everyone or whether it s just for union members We begin with a col lection of data CPSMar2004Extract wf1 taken from the March 2004 Current Population Survey We have data for about 100 000 individuals on wage rates measured in logs LNWAGE education ED age AGE and w
132. Page 335 Monday February 25 2013 10 06 AM 336 Chapter 14 A Taste of Advanced Estimation As a rough and ready adjust ment after looking at the resid ual plot we ll choose to give more weight to observations from 1952 on and less to those earlier We used a Stats By Classifica tion view of RESID to find error standard deviations for each sub period You can see that the residual standard deviation falls in half from 1952 We ll use this information to create a series ROUGH_W for weighting observations series rough_w 14 year lt 1952 6 year gt 1952 That s the heart of the trick in instructing EViews to do weighted least squares you need to create a series which holds the weight for every observation When per forming weighted least squares using the default settings EViews then multiplies each observation by the weight you supply Essentially this is equiv alent to replicating each observation in proportion to its weight EViews Illustrated book Page 336 Monday February 25 2013 10 06 AM Weighted Least Squares 337 The Weighted Option Open the least squares equation EQ01 in the workfile click the button and switch to the Options tab In the Weights group box select Inverse std dev from the Type dropdown and enter the weight series in the Weight series field Notice that we ve entered 1 ROUGH_W That s because 1 ROUGH_W is roughly propor tional to the inverse of the error st
133. Page 399 Monday February 25 2013 10 06 AM 400 Chapter 17 Odds and Ends Updates A Small Thing Objects in workfiles are somewhere between animate and inanimate If you open a graph view of a series and then change the data in the series the graph will update before your very eyes If you open an Estimation Output view of an equation and then change the data in one of the series used in the equation nothing at all will happen Some views update automatically and some don t Mostly the update or not decision reflects design guesses as to what the typical user would like to have happen To be sure estimates etc reflect the latest changes you made to your data redo the estimates Updates A Big Thing Quantitative Micro Software QMS posts program updates to www eviews com as needed Bug fixes are posted as soon as they become available Serious bugs are very rare and bugs outside the more esoteric areas of the program are very very rare QMS also posts free minor enhancements from time to time and once in a while posts updated documentation On occasion QMS releases a minor version upgrade 5 1 from 5 0 for example free of charge to current users Some of these minor upgrades include quite significant new features It pays to check the EViews website If you d like you can use the automatic updating feature to can check for new updates every day and install any available updates The auto matic update f
134. To add a regression line select Regression line from the Fit lines dropdown menu The default options for a regression line are fine so hit to dismiss the dialog We can see that the straight line gives a good rough description of how log volume moves over time even though the line doesn t hit very many points exactly The equation for the plotted line can be written algebraically as is the inter cept estimated by the computer and is the estimated slope Just looking at the plot we can see that the intercept is roughly 2 5 When LOGVOL looks to be about 4 Reaching back possibly to junior high school for the formula for the slope gives us an approximation for An eyeball approximation is that LOGVOL rises sixteen thousandths a bit over a percent and a half each quarter Estimating your first regression in EViews The line on the scatter diagram is called a regression line Obviously the computer knew the parameters and when it drew the line so backing the parameters out by eye may bring back fond memories but otherwise is unnecessarily convoluted Instead we turn now to regression analysis the most important tool of econometrics You can run a regression either by using a menu and dialog or by typing a command Let s try both starting with the menu and dialog method Pick the menu item Quick Estimate Equation at the top of the EViews window Then in the upper field type logvol c t Alternativel
135. a relatively high LSAT score Law School Admis sion Test we could fill out the Series Group for classify field with both WASH and LSAT gt 160 Now we get a table showing mean standard deviation and the number of observations for all four combinations of Washington resident not resident and high low LSAT The list of statistics reported appears in the upper left hand corner of the statistics table so that you ll have a key handy for reading the results The left hand side of the Statistics By Clas sification dialog has a series of checkboxes for selecting the statistics you d like to see The Output Layout field on the right hand side provides some control over the appearance of the table and whether you want margin statistics the All row and the All column Looking at statistics by classification makes sense when the classifying variable has a small set of distinct values When the classifying variable takes on a large number of values it s sometimes better to clump together values into a small number of groups or bins The Group into bins if field in the lower center of the dialog lets you instruct EViews to group different values of the classifying variable into a single bin See Binning Control later in this chapter EViews Illustrated book Page 202 Monday February 25 2013 10 06 AM Describing Series Picturing the Distribution 203 Describing Series Picturing the Distribution Som
136. ability distributions to a series and then plot the probability density You can supply the parameters of the distribution if you pre fer As an example we ve superimposed a normal distribution on top of the GPA histo gram EViews Illustrated book Page 206 Monday February 25 2013 10 06 AM Describing Series Picturing the Distribution 207 Empirical CDF Survivor and Quantile Graphs Just as a histogram or kernel density plot gives an estimate of the probability density func tion PDF a Cumulative Distribution plot presents an estimate of the cumulative distribu tion function CDF You may find it useful to think of the histogram and kernel density plots as graphical analogs to the Percent column in the one way tabulation shown above in One Way and the cumulative distri bution plot as the graphical ana log to the Cumulative Percent column Here s the CDF for GPA Survivor and Quantile plots pro vide alternative ways of looking at cumulative distributions The Distribution menu also pro vides links to a variety of Quan tile Quantile graphs and to a set of Empirical Distribution tests Quantile Quantile plots graph the empirical distribution of a series against a variety of theoretical probability distri butions e g normal uniform Empirical distribution tests provide corresponding formal tests of whether a series is drawn from a particular theoretical probability distribution For more on these topics s
137. ables in either static or dynamic forecast ing is not in any way affected by the sample period used for equation estimation g 2005m5 g2005m4 g 2005m5 g2005q5 g t 1 t 1 gt 1 EViews Illustrated book Page 228 Monday February 25 2013 10 06 AM Sample Forecast Samples 229 casts roll as far forward as you want assuming of course that the future values of the other right hand side series are also known Static Versus Dynamic in Practice Static versus dynamic forecasting are used to simulate answers to two different questions Suppose in the future you are going to be tasked with forecasting next month s currency growth When the date arrives you ll have all the necessary data to do a static forecast even though you don t have the data now Doing a static forecast now simulates the process you ll be carrying out later In contrast suppose you are going to be tasked with forecasting currency growth over the next 12 months When the date arrives you ll have to do a dynamic forecast Doing a dynamic forecast now simulates the process you ll be carrying out later One last practical detail You instruct EViews to do a static or dynamic forecast by picking the appropriate radio button in the Method field of the Forecast dialog Alternatively the command fit produces a static forecast and the command forecast pro duces as dynamic forecast as in forecast gf Sample Forecast Samples T
138. ably easy the mechanics are and then go over some of the more subtle issues more slowly Just Push the Forecast Button Our goal is to forecast the growth rate of currency in the hands of the public G You can find currency wf1 at the EViews website A line graph of the data in the series G appears to the right For this example we re going to model currency growth as a linear function of a time trend lagged currency growth and a different constant for each month of the year We need to estimate an equation for this model before we can make a forecast Here s the relevant command ls g trend g 1 expand month EViews Illustrated book Page 223 Monday February 25 2013 10 06 AM 224 Chapter 8 Forecasting The estimation results look fine Now to produce a forecast push the button When the Forecast dialog opens uncheck Forecast graph and Fore cast evaluation We ll talk about these later Set the Forecast sample at the lower right to 2000 through the end of the sample Hit You re done The forecasts for G are stored in the series GF EViews Illustrated book Page 224 Monday February 25 2013 10 06 AM Theory of Forecasting 225 To see how well we did let s plot actual and forecast currency growth together Pretty good forecasting no Perhaps leaning a little too heavily on seasonal fluctuations but basically pretty satisfactory You now know almost every thing you need to
139. ages which can be opened up and edited in EViews When you paste into your source document choose Paste Special and select Paste as EViews Object in the dialog To keep EViews completely in the picture you can paste as a link which will be updated whenever EViews changes Select Paste Link as EViews Object in the Paste Special dia log See the User s Guide for a complete discussion of using OLE Object Linking and Embed ding Exporting options are covered in some detail in Chapter 6 Intimacy With Graphic Objects but two options are used frequently enough that we mention them now By default EViews copies in color If the graph s final destination is black and white it s generally better to take a monochrome copy out of EViews because EViews makes different choices for line styles and backgrounds when it renders in monochrome The picture placed on the clipboard is provided as an enhanced metafile or EMF While this is the best all around choice some graphics and typesetting programs prefer encapsulated postscript or EPS This is particularly true for LaTeX and some desktop publishing pro grams Although EViews won t put an EPS picture on the clipboard it will save a file in EPS format using Proc Save graph to disk as described in the next section Hint Copying a graph object in the workfile window is different from copying from a graph window The latter puts a picture on the clipboard that can be pasted i
140. ainty about the true value of the regression coefficient The standard error of the regression abbreviated ser is the estimated standard deviation of the error terms In the inline display ser 0 967362 appears to the right of the regression equation proper EViews labels the ser as S E of regression reporting its value in the left column in the lower summary block Note that the third column of EViews regression output reports the standard error of the esti mated coefficients while the summary block below reports the standard error of the regres sion Don t confuse the two The final statistic in our scientific display is measures the overall fit of the regres sion line in the sense of measuring how close the points are to the estimated regression line in the scatter plot EViews computes as the fraction of the variance of the dependent variable explained by the regression See the User s Guide for the precise definition Loosely means the regression fit the data perfectly and means the regres sion is no better than guessing the sample mean The Pretty Important But Not So Important As the Last Section s Regres sion Results We re usually most interested in the regression coefficients and the statistical information provided for each one so let s continue along with the middle panel Hint EViews will report a negative for a model which fits worse than a model con sisting only of the sample
141. airs of beginning and ending dates In the illustration the pair 1 03 2005 1 07 05 specifies the first and last dates of the sam ple To pick out Monday and Wednesday through Friday specify the two pairs 1 03 2005 1 03 2005 1 05 2005 1 07 2005 Notice that we picked out a single date Monday with a pair that begins and ends on the same date EViews is clever about interpreting sample pairs as beginning and ending dates In a daily workfile specifying 2005m1 means January 1 if it begins a sample pair and January 31 if it ends a sample pair As an example smpl 2005m1 2005m1 picks out all the dates in January 2005 SMPLing the Sample To set the sample use the smpl command Not the related sample command which we ll get to in a second The command format is the word smpl followed by the sample you want used as in smpl 1 03 2005 1 03 2005 1 05 2005 1 07 2005 If you prefer the menu Quick Sample or the button which implements the smpl command not the sample command brings up the Sample dialog where you can also type in the sample The dialog is initial ized with the current sample for ease of editing SMPL Keywords Three special keywords help out in specifying date pairs first means the first date in the workfile last means the last date all means all dates in the workfile So two equiva lent commands are smpl all smpl first last Hint Above we used the word date For an un
142. al string variable value but rather in an object named ABC An evaluated string is a string variable name placed between squiggly braces as in 0 which tells EViews to use the name names or name fragment given by the string value For example if we enter age in the Run Pro gram dialog EViews will replace 0 in the program lines ls lnwage c 0 series e resid ls e 2 c 0 and then execute the following commands ls lnwage c age series e resid ls e 2 c age If we had entered UNION in the dialog the regressions would have been run on UNION instead of AGE Program Variables String variables live only while a program is being executed They aren t stored in the work file While they live you can use all the same string operations on string variables as you can on an alpha series EViews Illustrated book Page 383 Monday February 25 2013 10 06 AM 384 Chapter 16 Get With the Program In addition to string variables programs also allow control variables A control variable holds a number instead of a string and similarly lives only while the program is alive Con trol variables begin with an as in I String and control variables and string and scalar objects can be defined directly in a pro gram We ve written a slightly silly program which runs four regressions although the real purpose is to illustrate the use of control variables Running this progr
143. am is equivalent to entering the commands smpl if fe 0 ls lnwage c ed smpl if fe 1 ls lnwage c ed smpl if fe 0 ls lnwage c age smpl if fe 1 ls lnwage c age Hint A string variable in a program is a single string not one string per observation as in an alpha series Hint If you want to keep your string after the program is executed and save it in the workfile you should use put it into a string object Hint And if you want to keep your number after the program is executed and save it in the workfile you should use put it into a scalar object Hint When read aloud the exclamation point is pronounced bang as in bang eye EViews Illustrated book Page 384 Monday February 25 2013 10 06 AM Loopy 385 Loopy Program loops are a powerful method of telling EViews to repeatedly execute commands without you having to repeatedly type the commands Loops can use either control vari ables or string variables A loop begins with a for command with the rest of the line defining the successive values to be taken during the loop A loop ends with a next command The lines between for and next are executed for each specified loop value Most commonly loops with control vari ables are used to execute a set of commands for a sequence of numbers 0 1 2 In con trast loops with string variables commonly run the commands for a series of names that you supply Number loops The general form of the
144. ame You can also change the aspect ratio of the graph by click and dragging the bottom or right edges of the graph In contrast choosing 4 inches high and 3 inches wide gives a high and nar row frame The frame shape is mea sured in virtual inches What s really being deter mined is the width to height ratio and the font size relative to the frame size In addition these vir tual inches are used as the units of measurement for placing text determining margins etc So if you want to User position text half way across the frame you specify the x location as 4 inches in a and 1 5 inches in a frame One consequence of this is that changing the frame size may cause user positioned text to re locate itself 8 2 3 4 EViews Illustrated book Page 183 Monday February 25 2013 10 06 AM 184 Chapter 6 Intimacy With Graphic Objects Axes amp Scaling You may find that you visit the Axes amp Scaling section fre quently Its fea tures are both useful and very easy to use Assigning series to axes The Series axis assignment field on the Data scal ing page lets you assign each series to either the left or right axis with a radio button click or to the top and bottom axes for X Y Graphs This is espe cially important when graphing series with different units of measurement See Left and Right Axes in Group Line Graphs on page 140 in Chapter 5 Picture This
145. an war effort We can use shading to highlight interest rates during this period Define a shaded area by entering Left and Right obser vations in the Position field of the Lines amp Shading dialog Hint To put in multiple shaded areas use repeatedly Hint The Apply color to all vertical shaded areas checkbox lets you change the color of all the vertical shades on a graph with one command This checkbox morphs according to the type of line or shade selected For example if you ve set the combo for Orientation to Horizontal left axis the checkbox reads Apply color pattern amp width to all left scale lines EViews Illustrated book Page 171 Monday February 25 2013 10 06 AM 172 Chapter 6 Intimacy With Graphic Objects Shading can also be applied hori zontally After the Bretton Woods agreement on exchange rates broke down a number of Euro pean countries agreed to keep their exchange rates from rising or falling more than 2 25 percent Sometimes this worked and some times it didn t Shading visually highlights a band of the appropri ate width for the Belgian Dutch exchange rate in the figure to the right Lines Adding a vertical or horizontal line or lines to a graph draws the viewer s eyes to distinguish ing features For example long term interest rates are usually higher than short term interest rates This goes by the term called normal backwardation in case you were wondering The r
146. analysis While you can print lines typed in the command pane or save them to disk as COM MAND log point and click operations aren t recorded for posterity Entering com mands in a program and then running the program is the best way to create an audit trail in EViews Documentation Hint Anything to the right of an apostrophe in a program line is treated as a comment Writing lots of comments will make you happy later in life EViews Illustrated book Page 382 Monday February 25 2013 10 06 AM Program Variables 383 Instead of writing three separate programs we write one little program in which the right hand side vari able is replaced by an evaluated string variable argu ment In an EViews program A string variable begins with a sign holds text and only exists during program execution You may declare a string variable by entering the name an equal sign and then a quote delimited string as in y abc EViews automatically defines a set of string variables using arguments passed to the program in the Program arguments 0 1 field of the Run Program dialog The string variable 0 picks up the first string entered in the field The string variable 1 picks up the second string Etc You may refer to a string variable in a program using its name as in Y EViews will replace the string value with its string contents ABC In some settings we may be interested not in the actu
147. andard deviation As is generally true in EViews you can enter an expression wherever a series is called for Hint In fact if the weight is the EViews default scaling multiplies the data by the observation weight divided by the mean weight In theory this makes no difference but sometimes the denominator helps with numerical computation issues wi wi w EViews Illustrated book Page 337 Monday February 25 2013 10 06 AM 338 Chapter 14 A Taste of Advanced Estimation The weighted least squares esti mates include two summary statis tics panels The first panel is calculated from the residuals from the weighted regression while the second is based on unweighted residuals Notice that the unweighted from weighted least squares is a little lower than the reported in the original ordinary least squares estimate just as it should be Heteroskedasticity One of the statistical assumptions underneath ordinary least squares is that the error terms for all obser vations have a common variance that they are homoskedastic Vary ing variance errors are said in contrast to be heteroskedastic EViews offers both tests for heteroskedasticity and methods for producing correct standard errors in the presence of het eroskedasticity R2 R2 EViews Illustrated book Page 338 Monday February 25 2013 10 06 AM Heteroskedasticity 339 Tests for Heteroskedastic Residuals The Residual Diagnostics Heteroskedas
148. ant a series that assigned the mean education for women to women and the mean education for men to men This is accomplished with the Stats By family of functions meansby x y mediansby x y etc See the Command and Programming Reference for more func EViews Illustrated book Page 112 Monday February 25 2013 10 06 AM Relative Exotica 113 tions These functions summarize the data in X according to the groups in Y Optionally a sample can be used as a third argument Thus the command show fe ed meansby ed fe stdevsby ed fe shows gender and years of education followed by the mean and standard deviation of education for women if the individual is female and the mean and standard deviation of edu cation for men if the individual is male Expand the Dummies The expand function isn t really a data transformation function at all Instead expand x creates a set of temporary series One series is created for each unique value of X and the value of a given series is 1 for observations where X equals the corre sponding value For example expand fe creates two series in the command show fe expand fe If you give expand more than one series as an argument as in expand x y z series are created for all possible combinations of the values of the series The primary use of expand is as part of a regression specification where it generates a complete set of dummy variables Because it s often desirable t
149. aphs 157 CDF cumulative distribution function 207 cells 217 characters limiting number of 404 charts See graphs Chow tests 233 classification statistics by 201 210 testing by 212 213 coefficients constraining 357 location 27 significant 68 testing 67 75 78 360 vectors 342 343 color graphs 125 154 155 188 192 columns adjusting width 34 combining graphs 164 comma number separator 406 command pane 9 85 86 commands deprecated 86 EViews Illustrated book Page 415 Monday February 25 2013 10 06 AM 416 Index format 10 object 398 399 spaces in 63 64 use 9 See also specific commands comments 84 382 395 common samples 215 216 compression data 410 411 confidence interval forecasting 231 235 236 constant match average conversion 248 249 constant match sum conversion 249 contracting data 261 264 266 267 control variables 384 controls program 386 387 conversion data 244 246 251 408 text date 107 108 copying data 38 52 53 55 56 246 247 250 251 graphs 125 correlations 215 correlograms 318 319 321 count merges 262 covariances 215 creating audit trail 382 links 252 253 models 366 pages 240 242 programs 381 record of analyses 382 scenarios 376 series 8 10 27 28 85 113 302 303 system objects 357 VAR objects 361 workfiles 25 26 critical values 67 cross section data 275 275 cross section fixed effects 276 cross section identifiers 292
150. aphs and plots you re after If you like the more orderly approach continue on to the next chapter where we ll start the adventure of setting up your own workfile and entering your own data Now it s time to take a break for real EViews Illustrated book Page 22 Monday February 25 2013 10 06 AM Chapter 2 EViews Meet Data When you embark on an econometric journey your first step will be to bring your data into EViews In this chapter we talk about a variety of methods for getting this journey started on the right foot Unlike the blank piece of paper that appears metaphorically speaking when you fire up a word processor or the empty spreadsheet provided by a spreadsheet program the basic EViews document the workfile requires just a little bit of structuring information We begin by talking about how to set up a workfile Next we turn to manual entry typing data by hand While typing data is sometimes necessary it s awfully nice when we can just trans fer the data in from another program So a good part of the chapter is devoted to data import To get started here s an excerpt from the file AcadSalaries wf1 This file available on the EViews website excerpts data from a September 1994 article in Academe the journal of the American Association of University Professors The data give information from a survey of salaries in a number of academic disciplines The excerpt in Table 1 Academic Salary Data Excerpt sh
151. are days when computers are incredibly annoying The connection between pages is obvious to us but not to the computer because the observations aren t quite parallel in the two pages The infant mortality data includes an observation for the District of Columbia Reminder The identifier is the information that appears on the left in a spreadsheet view EViews Illustrated book Page 255 Monday February 25 2013 10 06 AM 256 Chapter 9 Page After Page After Page the revenue data doesn t The identifier for the former is a list of observation numbers 1 through 51 For the latter the identifier is observation numbers 1 through 50 Starting with Florida there s no identifier in common between the two series because Florida is observa tion 10 in one page and observation 9 in the other page Just matching observations by iden tifier won t work here Something more sophisticated is needed Matching through Links You can think of the match process as what computer scientists call a table look up Each time we need a value for REV we want EViews to go to the Revenue page and look up the value with the same state name in the Mortality page We understand that state is the meaningful link between the data in the two different pages We ll tell EViews to bring the data from the Revenue page into the Mortality page by creating a link and then filling in the Properties of the link with the information needed to make a match
152. ast values for volume storing them in the series VOLUMEF which now appears in the work file window Double click on Choose View Graph in the VOLUMEF window and select Line in the dialog to see the forecast values The first thing you ll notice is that nothing shows up on most of the plot We asked EViews to start the forecast in January 2001 and that s what EViews did so there is no forecast for most of our his torical period Click the button and enter 2000 LAST in the upper field of the Sample dialog Alternately use the slider bar to set the sample from 2000q1 to 2004q1 The graph snaps to a close up view of the last few years You have a volume forecast NYSE volume is forecast to rise over the forecast period from about 750 million shares to nearly 1 2 billion Mission accomplished What s Ahead This chapter s been a quick stroll through EViews just enough we hope to whet your appetite You can continue walking through the chapters in order but skipping around is fine too If you ll be mostly using EViews files prepared by others you might proceed to Chapter 3 Getting the Most from Least Squares to dive right into regressions to EViews Illustrated book Page 21 Monday February 25 2013 10 06 AM 22 Chapter 1 A Quick Walk Through Chapter 7 Look At Your Data for both simple and advanced techniques for describing your data or to Chapter 5 Picture This if it s gr
153. at the right and Total row at the bottom Looking at the right we see that perfect grades came from 0 61 of all applicants that 100 of applicants with perfect grades had perfect grades telling us that everything in the row is in the row which isn t very surprising and that 0 61 of this column had perfect grades which we already knew from the table in this cell Suppose that doing well on the LSAT and having good grades were independent We d expect that the percentage of students having both top grades and low test scores would be roughly the overall percentage having top grades times the overall percentage having low test scores For our data we d expect to fall into this cell In Hint The row and column percentages given in the Total column and row are some times called marginals because they give the univariate empirical distribution for grade and test score respectively So the row and column percentages correspond to the marginal probability distributions of the joint probability distribution described by the table as a whole 0 61 47 31 0 2886 EViews Illustrated book Page 218 Monday February 25 2013 10 06 AM Describing Groups Just the Facts Putting It Together 219 fact 0 31 do fall into this cell the difference might easily be due to random variation We could do the same calculation for all the cells and use this as a basis for a formal test of the hypothesis that grades and test scores a
154. at time series be stationary non explosive as opposed to nonstationary explosive This is an over simplification of some fairly complex issues But looking at a graph of LOGVOL it s clear that volume has exploded over time This suggests but doesn t prove that a time series model of the level of LOGVOL might be dicey A standard solution to this problem is to build a model of the first difference of the variable EViews Illustrated book Page 329 Monday February 25 2013 10 06 AM 330 Chapter 13 Serial Correlation Friend or Foe instead of modeling the level directly Given such a differenced model we then need to integrate the first differences to recover the levels So an ARMA model of the first differ ence is an AR Integrated MA or ARIMA model of the level Unit Root Tests A series that s stationary in first differ ences is said to possess a unit root EViews provides a battery of unit root tests from the View Unit Root Test menu For our purposes the default test suffices The User s Guide has an extended discussion of both EViews options and of the different tests avail able For this test the null hypothesis is that there is a unit root An excerpt of our test results are shown to the right Because the hypothesis of a unit root is not rejected we ll build a model of first differences Hint If you know etc then you can find by adding to You can find by adding and
155. ata will convert between monthly and quarterly data and will compute elapsed time between two observations in order to compute annualized rates of return EViews uses the id series to label all sorts of stuff from series windows for editing data through graphs of variables over time to recording the sample used for statisti cal estimation Notice how much easier it is to edit data with a meaningful date label right and how much more meaning you get out of a plot with the x axis labeled with a date rather than just an arbitrary observation number below Dated Irregular The Dated Irregular workfile structure stands in between the Dated Regular Frequency and the Unstructured Undated structures Each observation has a date attached but the obser vations need not be evenly spaced in time This sort of arrangement is especially useful for financial data where quotes are available on some days but not on others Tips for dating In addition to the Annual and Quarterly frequencies that we ve seen EViews offers a wide range of built in dated regular frequencies for those of you with dated data Multi year Annual Semi annual Quarterly Monthly Bimonthly Fort night Ten day Weekly Daily 5 day week Daily 7 day week Daily custom week Intraday and a special frequency Integer date which is a generalization of Unstructured Undated The Daily 5 day week and intra day frequencies are especially useful for Wall
156. ather than sequence num bers Rather than calling data for dentistry observation 1 and data for medicine observa tion 2 it might be a lot more meaningful to label them dentistry and medicine EViews lets us specify that one of the existing series obviously DISCIPLINE is the sensible choice should be used as the id series Hint EViews uses NA to indicate not available for numeric series and just an empty string for a not available alpha value The latter explains why the observations for DISCIPLINE are blank Hint To change the left to right order of series in a group use the menu View Group Members You ll see a list of series in the group Edit the text cut and paste is useful here re arranging the names into the desired order and click to accept the changes Alternatively there s no law against closing the group window and opening a new one Sometimes that s faster EViews Illustrated book Page 32 Monday February 25 2013 10 06 AM Identity Noncrisis 33 Changing the id series requires restructuring the workfile This is no big deal restructuring amounts essentially to telling EViews to use an id series Double click on Range in the upper pane of the workfile win dow or choose the menu item Proc Structure Resize Current Page Then choose Undated with ID series and fill in the series you want used for the id as illustrated here You ll notice that the Ra
157. ation is written and the next observation is written When there isn t any risk of confusion we sometimes drop the t The three observations might be writ ten Since typing subscripts is a nuisance lags and leads are specified in EViews by following a series name with the lag in parentheses For example if we have a series Hint Notice that is used both for comparison and as the assignment operator context matters Hint EViews generates a 1 0 as the result of a true comparison but only 0 is consid ered to be FALSE Any number other than 0 counts as TRUE So the value of the expression 2 AND 3 is TRUE i e 1 0 2 and 3 are both treated as TRUE by the AND operator Na Na Na EViews code for a number being not available is NA Arithmetic and logi cal operations on NA always produce NA as the result except for a few functions spe cially designed to translate NAs NA is neither true nor false it s NA 1 ONE_2_3 0 TWO_3_1 0 ONE_2_3 1 TWO_3_1 yt yt 1 yt 1 y 1 y y 1 EViews Illustrated book Page 87 Monday February 25 2013 10 06 AM 88 Chapter 4 Data The Transformational Experience named Y then Y 1 refers to the series lagged once Y 2 refers to the series lagged twice and Y 1 refers to the series led by one As an illustration the workfile 5Days wf1 contains a series Y with NASDAQ opening prices for the first five week
158. ations involving Japan and Italy are particularly high The menu View Residu als Covariance Matrix gives vari ances and covari ances instead of correlations Note that the variance of the residuals for Japan 347 8 is ten times the variance for Canada Generalized Least Squares and Heteroskedasticity Correction One of the assumptions underlying ordinary least squares estimation is that all observations have the same error variance and that errors are uncorrelated with one another When this assumption isn t true reported standard errors from ordinary least squares tend to be off and you forego information that can lead to improved estimation efficiency The tables above suggest that our pooled sample has both problems correlation across observations and differing variances EViews offers a number of options in the Weights menu in the Pool Estimation dialog shown to the right for dealing with heteroskedasticity We ll touch on a couple of them Relaxation hint This is a book about EViews not an econometrics tome If the title of this section just pushed past your comfort zone skip ahead to the next topic EViews Illustrated book Page 306 Monday February 25 2013 10 06 AM More Pool Estimation 307 Country Specific Weights To allow for a different variance for each country choose Cross section weights Compare these estimates to those we saw in Everyone Into the Pool May Not Be Fun on page 294 The estimated ef
159. base click on the Easy Query button to search the database for series using names descriptions or other characteristics Once you find the desired series highlight the series names and right mouse click or drag and drop to send the data into a new or existing workfile See the User s Guide for more detail EViews Illustrated book Page 245 Monday February 25 2013 10 06 AM 246 Chapter 9 Page After Page After Page Copy and Paste and Drag and Drop Let s decide to work at a monthly frequency One approach is to click on the Gdpc96 page tab to make it the active page then select RGDP and use the right click context menu to Copy Next click to activate on the Indpro page and paste Equivalently you can copy RGDP by drag and drop the RGDP icon onto the Indpro tab The icon will display a small plus sign when it is ready to be dropped EViews Illustrated book Page 246 Monday February 25 2013 10 06 AM Multiple Frequencies Multiple Pages 247 However you copy the data RGDP appears in the Indpro page not shown A dual scale graph gives a quick look at how the two series relate SIP and RGDP are measured in completely different units IP is simply an index set to 100 in 1997 RGDP is annualized and measured in billions of 1996 dol lars To find a conversion for mula we regress the latter on the former So when the industrial production number is announced each month we can get a sneak preview of real
160. bject An example is given in Simulating VARs in Chapter 15 Super Models Vector error correction cointegration tests structural VARs VARs have become an important tool of modern econometrics especially in macroeconom ics Since the User s Guide devotes an entire chapter to the subject we ll just say that the VAR object provides tools that handle everything listed in the topic heading above this para graph Quick Review A quick review of EViews advanced estimation features suggests that a year or two of Ph D level econometrics would help in learning to use all the available tools This chapter has tried to touch the surface of many of EViews advanced techniques Even this extended introduction hasn t covered everything that s available For example EViews offers a sophis ticated state space Kalman filter module As usual we ll refer you to the User s Guide for more advanced discussion EViews Illustrated book Page 364 Monday February 25 2013 10 06 AM Chapter 15 Super Models Most of EViews centers on using data to estimate something we d like to know often the parameters of an equation The model object turns the process around taking a model made up of linear or nonlinear possibly simultaneous equations and finding their solu tion We begin the chapter with the solution of a simple familiar model Next we discuss some of the ways that models can be used to explore different scenarios Of course
161. bles select the empty rectangle at the bottom that you want to fill in and choose Paste EViews does a very smart job of interpreting the data you ve copied and putting it in the right spot But if you find that Paste doesn t do just what you want try Paste Special which has extra options Sometimes the easiest way to combine observations from different sources is to read each source into a separate workfile create a master workfile with a range large enough to hold all your data and then manually copy from each small workfile into the master workfile Suppose that our dentistry and medicine data originated in one source and law agriculture and engineering in another We d begin by reading our data into two separate workfiles one with the first two observations and the other with the last three Looking at an excerpt from the two separate files we d see EViews Illustrated book Page 52 Monday February 25 2013 10 06 AM Adding Data To An Existing Workfile Or Being Rectangular Doesn t Mean Being Inflexible 53 We want to extend the range of the first workfile and then copy in the data from the second Select the workfile window for First Two wf1 Use File SaveAs to change the name to All Data Now double click on Range and change the range to 5 observations We need to be careful which workfile we re working in now Select the workfile window for Last Three wf1 Hit the button and Select All except C Resid
162. book Page 390 Monday February 25 2013 10 06 AM Appendix Sample Programs 391 next end of loop set sample to full sample smpl 1 100 show kernel density estimate for each coef freeze gra1 b1 kdensity draw vertical dashline at true parameter value gra1 draw dashline bottom rgb 156 156 156 beta1 show gra1 freeze gra2 b2 kdensity draw vertical dashline at true parameter value gra2 draw dashline bottom rgb 156 156 156 beta2 show gra2 Descriptive Statistics By Year Suppose that you wish to compute descriptive statistics mean median etc for each year of your monthly data in the series IP URATE M1 and TB10 One approach would be to link the data into an annual page see Chapter 9 Page After Page After Page and then compute the descriptive statistics in the newly created page Here s another approach which uses the statsby view of a series to compute the relevant statistics in two steps first create a year identifier series and second compute the statistics for each value of the identifier change path to program path path runpath cd path get workfile evworkfile data basics load evworkfile set sample smpl 1990 1 last create a series containing the year identifier series year year compute statistics for each year and freeze the output from each of the tables for var ip urate m1 tb10 name tab var EViews Illustrated book Page 391 Monday Fe
163. bruary 25 2013 10 06 AM 392 Chapter 16 Get With the Program freeze name var statby min max mean med year show name next More Samples These sample programs were all taken from online help found under Help Quick Help Ref erence Sample Programs amp Data Over 50 programs together with descriptions and related data are provided Reading the programs is an excellent way to pick up advanced tech niques EViews Illustrated book Page 392 Monday February 25 2013 10 06 AM Chapter 17 Odds and Ends An odds and ends chapter is a good spot for topics and tips that don t quite fit anywhere else You ve heard of Frequently Asked Questions Think of this chapter as Possibly Helpful Auxiliary Topics How Much Data Can EViews Handle EViews holds workfiles in internal memory i e in RAM as opposed to on disk Eight bytes are used for each number so storing a million data points 1 000 series with 1 000 observa tions each for example requires 8 megabytes Data capacity isn t an issue unless you have truly massive data needs perhaps processing public use samples from the U S Census or records from credit card transactions Current versions of EViews for 32 bit machines do have an out of the box limit of 4 million observa tions per series If you are working on a 64 bit machine you will be limited to 120 million observations per series How Long Does It Take To Compute An Estimate Probably not long enough f
164. bsite To create a new EViews series mea suring the length of two seconds type in the com mand pane series two_seconds one_second one_second EViews Illustrated book Page 83 Monday February 25 2013 10 06 AM 84 Chapter 4 Data The Transformational Experience Deconstructing Two Seconds Construction The results of the command are shown to the right They re just what one would expect Let s decon struct this terribly complicated example The basic form of the command is the command name series followed by a name for the new series followed by an sign followed by an algebraic expression A number of EViews cultural values are implicitly invoked here Let s go though them one by one Operations are performed on an entire series at a time In other words the addition is done for each observation at the same time This is the gen eral rule but we ll see two variants a little later one involving lags and one involving samples The sign doesn t mean equals it means copy the values on the right into the series on the left This is standard computer notation although not what we learned meant in school Note that if the series on the left already exists the values it contains are replaced by those on the right This allows for both useful commands such as series two_seconds two_seconds 1000 change units to milli seconds and also for some real
165. c of 2 corresponds approximately to a p value of 0 05 In the old days you d make the translation by looking at a t table in the back of a statistics book EViews just saves you some trouble by giving both t and p Not really about EViews digression Saying a coefficient is significant means there is statistical evidence that the coefficient differs from zero That s not the same as saying the coefficient is large or that the variable is important Large and important depend on the substantive issue you re working on not on statistics For example our estimate is that NYSE volume rises about one and one half percent each quarter We re very sure that the increase differs from zero a statement about statistical sig nificance not importance Consider two different views about what s large If you were planning a quarter ahead it s hard to imagine that you need to worry about a change as small as one and one half percent On the other hand one and one half percent per quarter starts to add up over time The estimated coefficient predicts volume will double each decade so the estimated increase is certainly large enough to be important for long run planning EViews Illustrated book Page 68 Monday February 25 2013 10 06 AM The Pretty Important But Not So Important As the Last Section s Regression Results 69 We talked above about scientific conventions for reporting results and sho
166. cal Graphs 161 Multiple Series as Factors Having multiple series in a graph is sort of like hav ing multiple categories for a single series in that there s more than one group of data to graph EViews recognizes this Use the Treat multiple series in this Group object as menu to treat the series the series in this group were WSAL_VAL and HRSWK as a factor If we d set Treat multiple series in this Group object as First within factor both WSAL_VAL and HRSWK would appear in the same plot as shown below Because of the dif ference in scales for the two series First within factor wouldn t be a sensible choice for this application Hint EViews will produce graphs with as complex a factor structure as you d like That doesn t make complex structures a good idea Anything much more complicated than the graph above starts to get too complicated to convey a clear visual message EViews Illustrated book Page 161 Monday February 25 2013 10 06 AM 162 Chapter 5 Picture This Polishing Factor Layouts For categorical graphs the Graph Type group on the left hand side of the dialog includes a Categorical options section with a number of fine tuning options We dis cuss the most used options here leav ing the rest to your experimentation and to the User s Guide Distinguishing Factors In the graph depicted earlier all the bars are a single color and pattern Contrast the graph to the graph shown below
167. can add a for command with a string vari able Other Program Controls In addition to the command form used throughout EViews Illustrated commands can also be written in object form In a program use of the object form is often the easiest way to set options that would normally be set by making choices in dialog boxes Sometimes you don t want the output from each command in a program For example in a Monte Carlo study you might run 10 000 regressions saving one coefficient from each regression for later analysis but not otherwise using the regression output A Monte Carlo study is used to explore statistical distributions through simulation techniques Creating 10 000 equation objects is inefficient and can be avoided by using the object form For example the program for i 1 to 10000 ls y c x next runs the same regression 10 000 times creating 10 000 objects and opening 10 000 win dows Actually you aren t allowed to open 10 000 windows so the program won t run The program equation eq for i 1 to 10000 Hint Loops are much easier to read if you indent EViews Illustrated book Page 386 Monday February 25 2013 10 06 AM A Rolling Example 387 eq ls y c x next uses the object form and creates a single object with a single re used window EViews offers other programming controls such as if else endif while loops and sub routines We refer you to the Command and Programming Refere
168. can also automate repetitive commands by using numerical and string for next loops An EViews program is an excellent way to document your operations and compared to manually typing every command can save a heck of a lot of time Nearly all operations can be written as commands suitable for inclusion in a program Sim ple loops are quite easy to use A limited set of matrix operations are available for more complex calculations However it s probably best to think of the program facility as provid ing a very sophisticated command and batch scripting language rather than a full blown programming environment And if you come across a need for something not built in see the no not the User s Guide this time see the 900 page Command and Programming Reference Appendix Sample Programs Rolling forecasts The dynamic forecast procedure for an equation produces multi period forecasts with the same set of estimated parameters Suppose instead that you want to produce multi period forecasts by reestimating the parameters as new data become available To be more specific suppose you want to produce forecasts up to 4 periods ahead from each subsample where each subsample is moved 4 periods at a time Hint If you want to learn more about writing EViews programs start with Chapter 17 EViews Programming in the User s Guide EViews Illustrated book Page 388 Monday February 25 2013 10 06 AM Appendix Sample Programs 389
169. cast should only have difficulty at the beginning of the forecast period since thereafter actual lagged G picks up non anom alous data Using this new estimate shown above to the right as a basis for a static forecast through the HEINLEIN period we can set the Forecast dia log to save a new forecast to plot the forecast and confidence intervals and to show us a new forecast evalu ation R2 EViews Illustrated book Page 234 Monday February 25 2013 10 06 AM Forecasting Beneath the Surface 235 When the Forecast graph and the Forecast evaluation options are both checked EViews puts the confidence interval graph and statistics together in one window We ve definitely gotten a bit of improvement from chang ing the sample One last plot The seasonal forecast swings are still some what larger than the actual sea sonal effects but the forecast is really pretty good for such a simple model Forecasting Beneath the Surface Sometimes the variable you want to forecast isn t quite the variable on the left of your esti mating equation There are often statistical or economic modeling reasons for estimating a transformed version of the variable you really care about Two common examples are using logs rather than levels of the variable of interest and using first differences of the variable of interest EViews Illustrated book Page 235 Monday February 25 2013 10 06 AM 236 Chapter 8 Forecasting In the
170. ce page Do a contraction using the regular mean method Finally exponentiate the resulting series in the destination page as in geo_av exp lnx Sometimes as in this example this sort of work around is easy sometimes it isn t Two Hints and A GotchYa In our examples the Source ID and Destination ID each specified a single series with the same name It s perfectly okay to have different names in these fields EViews matches according to the values found in the respective series What s more you can put multiple series in both fields If we entered AA BB CC for the source and One Two Three for the destination EViews would match observations where the value of AA matched the value of ONE and the value of BB matched the value of TWO and the value of CC matched the value of THREE Normally observations with NA in any of the Source ID or Destination ID series are tossed out of matches Check the checkbox Treat NAs as ID Category to tell EViews to treat NA as a valid value for matching And then the gotchya risk When you paste by value the matching and merging is done right away When you use a link the matching and merging is re executed each time a value of the link is called for Remember that the link specification has a sample built into it and that this sample is re evaluated each time the link is recomputed If the observations included in this sample are changing be sure that the change is as you intended S
171. cepts of objects object commands and object views No everything can be done by combining point and click with typing the straightfor ward commands that we ve used throughout EViews Illustrated But if you prefer doing everything via the command line or if you want an exact record of the commands issued you may find the fine grain control offered by objects helpful The Command and Programming Reference includes extensive tables documenting each object type and the commands and views associated with each Workfile Backups When you save a workfile EViews keeps the previous copy on disk as a backup changing the extension from wf1 to f1 For example the first time you save a workfile named foo the file is saved as foo wf1 The second time you save foo the name of the first file is changed to foo f1 and the new version becomes foo wf1 The third time you save foo the first file disappears the second incarnation is changed to foo f1 and the third version becomes foo wf1 It s okay to delete backup versions if you re short on disk space On the other hand if some thing goes wrong with your current workfile you can recover the data in the backup ver sion by changing the backup filename to a name with the extension wf1 For example to read in foo f1 change the name to hope_this_saves_my_donkey wf1 and then open it from EViews EViews Illustrated book
172. cific coefficients Regressors and AR terms The Options tab provides a variety of methods for robust estimation of standard errors and also options for controlling exactly how the estimation is done These are of course discussed at length in the User s Guide Nomenclature hint SUR stands for Seemingly Unrelated Regression EViews Illustrated book Page 308 Monday February 25 2013 10 06 AM Getting Data In and Out of the Pool 309 Getting Data In and Out of the Pool Since pooled series are just ordinary series you re free to load them into EViews any way that you find convenient But there are two data arrangements that are common unstacked and stacked These data arrangements correspond to the spreadsheet arrangements we saw in Spreadsheet Views earlier in the chapter Unstacked data are read through a standard File Open See Chapter 2 EViews Meet Data EViews provides some special help for stacked data Importing Unstacked Data Here s an excerpt of an Excel spreadsheet with unstacked data Since we have ordinary series with conveniently chosen names load in the spreadsheet in the usual way create a pool object with suffixes CAN FRA etc and bob s your uncle Importing Stacked Data The Direct Method Here s an excerpt of an Excel spreadsheet with stacked data stacked by cross section We ve hid den some of the rows so you can see the whole pat tern All the data for the first country appear
173. come accounting iden tity it should really be marked as an identity Click the Identity radio button on the lower right of the dialog and then to return to the equations view To complete the model we need to bring in the consumption function The estimated consumption function is stored in the workfile as Select this equation in the workfile window and then copy and paste or drag and drop it into the model window EViews checks to be sure that you really want to link the estimated consumption function into the model Since you do choose The model is now complete EViews Illustrated book Page 367 Monday February 25 2013 10 06 AM 368 Chapter 15 Super Models Solving the Model So what s the homework answer Click The Model Solution dialog appears with lots and lots of options We re not doing anything fancy so just hit EViews will find a numerical solution for the simultaneous equation model we ve specified Two windows display new information The model solu tion messages window below right provides information about the solution The workfile window below left has acquired two series ending with the suffix __0 The model window gives details of the solution technique used Complicated models can be hard for even a computer to solve but this model is not complex so the details aren t very interesting Note at the bottom of the window that EViews solved the model essentially insta
174. ct the pool definition in any way Making Groups As you ve seen there are lots of ways to manipulate a group of pooled series from the pool window But sometimes it s easier to include all the series in a standard EViews group and then use group proce dures Plotting pooled series is one such example Choose Make Group from the Proc menu and enter the series you want in the Series List dialog An untitled group will open If we like we could make a quick plot of Y for the pooled series by switching the group to a graph view See Chapter 5 Picture This Hint genr is a synonym for the command series in generating data Hence the name on the button PoolGenr EViews Illustrated book Page 303 Monday February 25 2013 10 06 AM 304 Chapter 12 Everyone Into the Pool More Pool Estimation You won t be surprised that estimates with pools come with lots of interesting options We touch on a few of them here For complete information see the User s Guide Residuals Return to the first pooled estimate in the chapter the one with a common intercept for all series Pick the menu View Residuals Graphs The first thing you ll notice is that squeez ing six graphs into one window makes for some pretty tiny graphs Residuals are supposed to be centered on zero The second thing you ll notice is that the residuals for Canada are all strongly positive those for Germany are mostly positive and Italy s res
175. d In fact there s an important general rule about the evaluation of sample specifications The sample is re evaluated every time EViews processes data The example at hand includes the clause if y gt 2100 Every time the values in Y change the set of points included in the sample may change For example if you edit Y changing an observation from 2200 to 2000 that observation drops out of sample S1 Three Simple SAMPLE amp SMPL Tricks To save the current sample specification give the command sample or use the menu Object New Object and pick Sam ple Either way the Sample dialog opens with the current workfile sample as initial values Immediately hit to save the current specification in the newly defined sample object Remember that the if clause in a sample includes observations where the logical condition evaluates to TRUE 1 and excludes observations where the condition evaluates to False 0 Hint In a sample if clause NA counts as false EViews Illustrated book Page 98 Monday February 25 2013 10 06 AM Simple Sample Says 99 Freezing the current sample To freeze a sample so that you can reuse the same observations later even if the variables in the sample specification change first create a variable equal to 1 for every point in the current sample series sampledummy 1 Then set up a new sample which selects those data points for which the new variable equals 1 sampl
176. d from EViews Sometimes trouble can be avoided or at least mitigated by changing the graphics default through the menu Options Graphics Defaults clicking the Exporting section and setting Text labels to Keep label as a single block of text so that it can be edited in other programs Unfortunately even this isn t guaranteed to work with one very popular Word processing program The moral is try to completely polish your graphic in EViews before exporting it Hint The six custom text items in the illustration were added by clicking six times Hint The Text Labels dialog is for the text you add to the graphic Text placed by EViews such as legends and axis labels is adjusted through the Graph Options dialog Uh except for the occasional automatically generated title which is tweaked through Text Labels Not to worry double click on text and the appropriate dialog opens EViews Illustrated book Page 170 Monday February 25 2013 10 06 AM Shady Areas and No Worry Lines 171 Shady Areas and No Worry Lines Adding shaded areas or vertical or horizontal lines to graphs is a very effective way of focus ing your audience s attention on specific aspects of a plot The button brings up the Lines amp Shading dialog which is used to place both lines and shades Shades Beginning in 1942 and ending with the March 4 1951 Treasury Federal Reserve Accord the Fed eral Reserve kept interest rates low in support of the Americ
177. d least squares problem which opened the chapter This time we ll estimate the variances and coefficients jointly Our first step is to create a new LogL object using either the Object New Object menu or a command like logl weighted_example Opening brings up a text area for entering definitions Think of the com mands here as a series of series commands only without the command name series being given that EViews will execute in sequential order These commands build up the definition of the contribution to the likelihood function The file also includes one line with the keyword logl which identifies which series holds the contribution to the likelihood function EViews Illustrated book Page 355 Monday February 25 2013 10 06 AM 356 Chapter 14 A Taste of Advanced Estimation Let s take apart our example specifi cation shown to the right We broke the definition of the error term into two parts simply because it was easier to type The first equation defines the seasonal component The second equation is the differ ence between observed currency growth and predicted currency growth The third equation defines the error term standard deviation as coming from either the early period variance C 15 or the late period variance C 16 Note that all these definitions depend on the values in the coefficient vector C and will change as EViews tries out new coefficient values The fourth line defines LOGL1 which give
178. d pane to disk default file name com mand log by clicking anywhere in the pane and choosing File Save or File SaveAs Some folks have a taste for using menus rather than typing commands We could have created TWO_SECONDS with the menu Quick Gener ate Series Using the menu and Generate Series by Equation dialog has the advantage that you can restrict the sample for this one operation without changing the workfile sam ple More on samples in the next section There s a small disadvantage in that unlike when you type directly in the command pane the equation doesn t appear in the command pane so you re left without a visual record Obvious Operators EViews uses all the usual arithmetic operators Operations are done from left to right except that exponentiation comes before multiplication and division which come before addition and subtraction Numbers are entered in the usual way 123 or in scientific notation 1 23e2 EViews handles logic by representing TRUE with the number 1 0 and FALSE with the num ber 0 The comparison operators gt lt gt greater than or equal lt less than or equal and lt gt not equal all produce ones or zeros as answers Deprecatory hint Earlier versions of EViews used the command genr for what s now done with the distinct commands series alpha and f
179. date in the work file EViews accepts a variety of conventions for writing a particular date 2001 1 means the first period of 2001 and 2001q1 more specifically means the first quarter of 2001 Since the periods in our data are quarterly the two are equivalent EViews Illustrated book Page 7 Monday February 25 2013 10 06 AM 8 Chapter 1 A Quick Walk Through Generating a new series Let s turn to another approach to thinking about trading volume Our first line graph before we shortened the sample presented a picture which looks a lot like exponential growth over time A standard trick for dealing with exponential growth is to look at the logarithm of a variable relying on the identity In order to look at the trend in the log instead of in the level we ll create a new variable named LOGVOL which equals the log of VOLUME This can be done either with a dialog or by typing a command We ll do the former first Choose the menu item Quick Generate Series to bring up a dialog box In the upper field type logvol log volume Notice that in the lower field the sample is still set to use only the 21st century part of our data This matters as we ll see in a moment The workfile window now has a new object Hint The new sample applies to all our work until we change it again not just this one graph Note the change in the Sample line of the workfile window y egt y log gt EVi
180. dated workfile substitute observa tion number To pick out the first ten observations in an undated workfile use the pair 1 10 EViews Illustrated book Page 96 Monday February 25 2013 10 06 AM Simple Sample Says 97 Arithmetic operations are allowed in specifying date pairs For example the second date in the workfile is first 1 To specify the entire sample except the first observation use smpl first 1 last The first ten and last ten observations in the workfile are picked by smpl first first 9 last 9 last Smpl Splicing You can take advantage of the fact that observations outside the current sample are unaf fected by series operations to splice together a series with different values for different dates For example the commands smpl all series prewar 1 smpl 1945m09 last prewar 0 smpl all first create a series equal to 1 0 for all observations Then it sets the later observations in the series to 0 leaving the pre war values unchanged SMPLing If A sample specification has two parts both of which are optional The first part the one we ve just discussed is a list of starting and ending date pairs The second part begins with the word if and is followed by a logical condition The sample consists of the observations which are included in the pairs in the first part of the specification AND for which the logi cal condition following the if is true If no date pair
181. days of 2005 Looking at Tuesday s data you ll see that the value for Y 1 is Monday s opening price and the value for Y 1 is Wednesday s opening price Y 1 for Monday and Y 1 for Friday are both NA because they represent unknown data the opening price on the Friday before we started collecting data and the opening price on the Monday after we stopped collecting data respectively The group shown above was created with the EViews command show daynames y y 1 y 1 If we wanted to compute the percentage change from the previous day we could use the command series pct_change 100 y y 1 y 1 The Entire Series At A Time Exception For Lags A couple of pages back we told you that EViews operates on an entire series at a time Lags are the first exception When the expression on the right side of a series assignment includes lags EViews processes the first observation assigns the resulting value to the series on the left and then processes the second observation and so on The order matters because the assignment for the first observation can affect the processing of the second observation Consider the following EViews instructions ignoring the smpl statements for the moment smpl all series y 1 smpl first 1 last y y 1 5 Hint In a regularly dated workfile 5Days wf1 for example one lag picks up data at In an undated or an irregularly dated workfile one lag simply picks up the
182. der type dropdown gives options that can handle most com mon file arrangements In this case choose Names in last line and away we go Names in last line means the names are at the end of the header information right before the data begins a pretty common arrangement Hint EViews examines your spreadsheet and generally makes a pretty intelligent guess about which part of the spreadsheet you d like read You can also set the range manu ally in the Spreadsheet read dialog You ll find it saves time if you define a named range demarcating your data in Excel In this way you need only select the named range when EViews reads in the spreadsheet Hint There s no harm in trying out EViews first guess If the results aren t what you re looking for throw them out re open the file and set the controls the way you want in the Spreadsheet read dialog EViews Illustrated book Page 42 Monday February 25 2013 10 06 AM The Import Business 43 Being Date Savvy It s not at all unusual for a data file to include the date of each observation EViews does a surprisingly good job of guessing that a particular column of data consists of dates that ought to be used as identifiers in the workfile Here s an excerpt from an Excel file NZ Unemploy ment xls downloaded from Statistics New Zealand The data are quarterly unemployment but note that each obser vation is labeled with the last month of the quarter and the year
183. e s a version of our debt graph using spikes to show the first quarter of each year with padding for excluded obs Seasonal Graphs The standard line graph to the right shows U S retail and food service sales over a dozen years Notice the regular spikes How fast can you say Christmas EViews Illustrated book Page 134 Monday February 25 2013 10 06 AM Picture One Series 135 Change the Graph type to Seasonal and the right hand side of the Graph Option dialog changes to give you choices of two kinds of seasonal graphs Paneled lines amp means draws one line graph for each season and also puts in a horizontal line to mark the sea sonal mean Since our retail sales data Retail Sales wf1 is in a monthly workfile that means twelve lines Using a Paneled lines amp means graph it s easy to see that December sales are relatively high and that sales in January and February are typically low EViews Illustrated book Page 135 Monday February 25 2013 10 06 AM 136 Chapter 5 Picture This Multiple overlayed lines graphs also provide one line for each season but use a common date axis For our retail sales data the Multiple overlayed lines graph does a particularly good job of showing how December higher and January and Febru ary lower compare to the remaining months Distribution Quantile Quantile and Boxplots Distribution graphs quantile quantile plots and boxplots provide p
184. e Boxplots in Chapter 7 Look At Your Data For more information see the User s Guide EViews Illustrated book Page 191 Monday February 25 2013 10 06 AM 192 Chapter 6 Intimacy With Graphic Objects Quick Fonts In the Quick Fonts page you can easily set the font and font size globally for all axes text objects observa tion labels and or the leg end Use caution with this quick and easy method it can not be undone for text objects so be sure of your edits before clicking Apply Objects The Object options page controls the style but not the content for lines shading and text objects You can set the style for a given object directly in the Text Labels and Lines amp Shading dialogs The Object options page lets you set the default styles for any new objects in the graph at hand You can also change the styles for existing objects in the graph by checking the relevant Apply to existing box EViews Illustrated book Page 192 Monday February 25 2013 10 06 AM The Impact of Globalization on Intimate Graphic Activity 193 Graph Updating The Graph Updating section lets you specify if you would like your graph to update with changes in the underlying data If you select Manual or Auto matic the bottom half of the page becomes active where you may specify the update sample The Impact of Globalization on Intimate Graphic Activity If you do lots of similar graphs you
185. e arranged in fixed columns as is the case in this excerpt try the Fixed width fields option described in the next section Hint If you have a choice get your data tab delimited Life is better with tab EViews Illustrated book Page 45 Monday February 25 2013 10 06 AM 46 Chapter 2 EViews Meet Data Fixed width fields Another very popular for mat especially with older data is to skip the issue of delimit ers entirely and put each data field in fixed columns DIS CIPLINE might be in columns 1 through 18 NONACADSAL in columns 19 through 360 etc If you choose the radio button Fixed width fields in the ASCII Read dialog EViews will show you its best guess as to where fields end In this exam ple EViews guess isn t quite right Hit so that you can drag the col umn dividers to the right loca tions EViews Illustrated book Page 46 Monday February 25 2013 10 06 AM The Import Business 47 Now manually adjust the col umns After you have the column bound aries where they belong hit Explicit format EViews provides a third very powerful option for describing the layout of data You can pro vide an Explicit Format which can be specified in EViews notation or using notation from either of two widely used computer languages Fortran format notation or C scanf notation See the User s Guide for more information EViews reads about two dozen other file formats including files
186. e frozensample if sampledummy Now any time you give the command smpl frozensample you ll restore the sample you were using Creating dummy variables for selected dates To create a dummy zero one variable that equals one for certain dates and zero for others first save a sample specification including the desired dates Later you can include the sam ple in a series calculation taking advantage of the fact that in such a calculation EViews evaluates the sample as 1 for points in the sample and 0 for points outside Try deconstruct ing the following example sample s2 first 1 last if y y 1 smpl all series sameyasprevious y s2 999 1 s2 Got it The first line defines S2 as holding a sample specification including all observations for which Y equals its own lagged value The second line sets the sample to include the entire workfile range The third line creates a new series named SAMEYASPREVIOUS which equals Y if Y equals its own lagged value and 999 otherwise The trick is that the sample S2 is treated as either a 1 or a 0 in the last line Nonsample SMPLs Each workfile has a current sample which governs all operations on series except when it doesn t In other words some operations and commands allow you to specify a sample which applies just to that one operation For these operations you write the sample specifi cation just as you would in a smpl command In some cases the string defining the sample need
187. e or LSAT scores might still be of interest for example There s no right or wrong about this side effect You just want to be aware that it s happening The remainder of the statistics panel reports characteristics of the data sample mean median etc Hint If you want to eliminate data errors for one series without affecting which obser vations are used for other series in an analysis change the erroneous values to NA instead of cutting them out of the sample EViews Illustrated book Page 198 Monday February 25 2013 10 06 AM Describing Series Just The Facts Please 199 The statistic at the bottom of the panel the Jarque Bera tests the hypothesis that the sample is drawn from a normal distribution The statistic marked Probability is the p value asso ciated with the Jarque Bera In this example with a p value of 0 000 the report is that it is extremely unlikely that the data follows a normal distribution One Way To look at the complete distribution of a series use One Way Tabulation which lets you Tabulate Series Initially it s best to uncheck both Group into bins if checkboxes Eliminating binning ensures that we see a complete list of every value appearing in the series from low to high as well as a count and cumulative count of the number of observations taken by each value Export Hint If you double click on the statistics panel the Text Labels dialog opens This is the place to manipula
188. e the appearance of codes into something more pleasant to read The codes them selves are unchanged it s their appearance that s improved To create a new value map named GENDER use Object New Object ValMap or type the command valmap gender The ValMap editor view pops up showing two columns Values 0 or 1 in this case go on the left and the corresponding labels male female go on the right Two default mappings appear at the top blanks and NAs are dis played unchanged EViews Illustrated book Page 109 Monday February 25 2013 10 06 AM 110 Chapter 4 Data The Transformational Experience To fill out the value map enter a list of codes on the left and labels on the right When you close the valmap window is stored in the workfile with an icon that says map The values can be either numeric or text and there s no harm in providing maps for values that aren t used To tell EViews to use the map just cre ated open the series FE click the button choose the Value Map tab and type in GENDER Displayed FE now uses much friendlier labels EViews will use the new labels wherever possible For example the command ls lnwage ed expand fe Hint If you want to see FE s underlying codes switch the display type menu in the series spreadsheet view from Default to Raw Data EViews Illustrated book Page 110 Monday February 25 2013 10 06 AM What Are Your Values 111 which regresses a
189. e two Califor nia entries EViews would average the two entries without any warning We could instead specify No contractions allowed which instructs EViews to copy the relevant value but to display an error message if it finds more than one entry for a state Unique values is almost the same as No contractions allowed except that if all the values for a category are identi cal the link proceeds In other words if we had entered California twice with a unionization value of 0 0303 Unique values would proceed while No contractions allowed would fail You can t provide your own contraction method but you may be able to construct a work around EViews doesn t provide for contraction by geometric average for example but you can roll your own A geometric average is defined as Hint As usual EViews knows more than one way to skin a cat If we hadn t had any interest in seeing the state by state averages we could have used the MEANSBY function discussed briefly in Chapter 4 Data The Transformational Experience The following command would produce the same results as we ve just seen ls lnwage c ed age union meansby union gmstcen if not isna lnwage Q Can I specify my own function in place of one of the built in contraction methods A No A Mostly no EViews Illustrated book Page 266 Monday February 25 2013 10 06 AM Two Hints and A GotchYa 267 To do this by hand define a series lnx log x in the sour
190. ead directly by most word processors Select Redirect in the Print dialog and enter a file name in the Filename field As shown you ll end up with results stored in the file some results rtf Right click and choose Select non empty cells or hit Ctrl A it s the same thing Copy and then paste into a word proces sor Freeze it If you have output that you want to make sure won t ever change even if you change the equation specification hit Freezing the equation makes a copy of the current view in the form of a table which is detached from the equation object The original equation is unaffected You can then this frozen table so that it will be saved in the workfile See Chapter 17 Odds and Ends Hint Before saving the file switch to the equation s label view and write a note to remind yourself why you re using this equation EViews Illustrated book Page 71 Monday February 25 2013 10 06 AM 72 Chapter 3 Getting the Most from Least Squares Summary Regression Statistics The bottom panel of the regres sion provides 12 summary statis tics about the regression We ll go over these statistics briefly but leave technical details to your favorite econometrics text or the User s Guide We ve already talked about the two most important numbers R squared and S E of regression Our regression accounts for 85 percent of the variance in the dependent vari able and the e
191. easier to look at than our graph of the original VOLUME variable We might conclude from looking at our LOGVOL line graph that NYSE volume rises at a more or less constant percentage growth rate in the long run with a lot of short run fluctua tion Or perhaps the picture is better represented by slow growth in the early years a drop in volume during the Great Depression starting around 1929 and faster growth in the post War era We ll leave the substantive question as a possibility for future thought and turn now to building a regression model and making a forecast of future trading volume Looking at a pair of series together Our line graph goes quite far in giving us a qualitative understanding of the behavior of vol ume over time For a quantitative understanding we d like to put some numbers to the upward trending picture of LOGVOL If we have a variable t representing time 0 1 2 3 then we can represent the idea of an upward trend with the algebraic model where the coefficient gives the quarterly increase in LOGVOL Hint Menu items both in the menu bar at the top of the screen and menus chosen from the button bar change to reflect the contents of the currently active window If the menu items differ from those you expect to see the odds are that you aren t look ing at the active window Click on the desired window to be sure it s the active win dow volumet log a bt b EViews Illustrat
192. eate EViews programs in your favorite text editor or any word processor able to save standard ASCII files EViews Illustrated book Page 381 Monday February 25 2013 10 06 AM 382 Chapter 16 Get With the Program As a practical matter you probably don t want to run the same regression on the same series over and over and over On the other hand you might very well want to apply the same data transforma tion to a number of different data sets For exam ple the Current Population Survey CPS records various measures of educational attainment but doesn t provide a years of education variable The program transformcps_ed prg translates the series A_HGA that appears in the data supplied in the CPS into years of education ED When new CPS data are released each March we can run transformcps_ed again to re create ED You Want To Have An Argument You might want to run the same regression over and over again on different series Our first program regressed LNWAGE on ED and then looked to see if there is a relation between the squared residuals and ED Suppose we want to execute the same procedure with AGE and then again with UNION as the right hand side variable Hint If you choose the Quiet radio button in the Run Program dialog not shown here see below programs execute faster and more peacefully But Verbose is some times helpful in debugging Documentary Hint It s sometimes important to keep a record of your
193. eature can be enabled or disabled from the main Options menu under the Etymological Hint Foo is a generic name that computer types use to mean any file or any variable Foo is a shortened version of foobar Foobar derives from the World War II term fubar which itself is an acronym for Fouled Up Beyond All Rec ognition or at least that s the acronymization given in family oriented books such as the one you are reading The transition from fubar to foobar is believed to have arisen from the fact that if computer types could spell they wouldn t have had to give up their careers as English majors For further information contact the Professional Organization of English MajorsTM EViews Illustrated book Page 400 Monday February 25 2013 10 06 AM Ready To Take A Break 401 EViews Auto Update from Web menu item Or manually check for updates from by select ing Check now or by selecting EViews Update from the Help menu Ready To Take A Break If EViews is taking so long to compute something that you d like it to give up hit the Esc Escape key EViews will quit what it s doing and pay attention to you instead Help It is conceivable that you ve read EViews Illustrated and nonetheless may someday need more help The Help menu has all sorts of goodies for you Complete electronic versions of the User s Guide Command and Programming Reference and Object Refe
194. ect a graph object which is independent from the original series or group you were looking at A graph object is fundamentally a picture that happens to have started life as a graphic view You make a graph object by looking at a graphic view as we are at the moment and hitting the button A dialog opens allowing you to choose how you want your new graph object to be tied to the underlying data Selecting Manual or Automatic update will keep the graph object tied to the data in the series or group that it came from When the data changes the graph object will reflect the new values To update the graph with any applicable changes select Automatic To control when the graph update occurs select Manual If you d rather freeze the graph as a snapshot of its current state select Off Click to create the graph object Hint If you select Off and then decide you d like to relate the graph object to its under lying data again you can always change your selection later in the master Graph Options dialog in the Graph Updating section EViews Illustrated book Page 120 Monday February 25 2013 10 06 AM A Simple Soup To Nuts Graphing Example 121 A new window opens with the same picture but with Graph in the titlebar instead of Group and with a different set of buttons in the button bar Named graph objects appear in the workfile window with a icon An orange icon alerts us to a graph that will update with cha
195. ect Scenario tab Scenario 1 is associated with the alias _1 If you like you can rename the new scenario to something more mean ingful or change the suffix but we ll just click for now We want to instruct EViews to use different values of G in this sce nario We ll create the series G_1 with the command series g_1 g 10 Overriding Baseline Data Back in the model window click and then right click on G and choose Properties Hint We chose the name G_1 because the suffix has to match the scenario alias EViews Illustrated book Page 376 Monday February 25 2013 10 06 AM Your Second Homework 377 On the Properties dialog check Use override in series in scenario to instruct EViews to substi tute G_1 for G Now New series Y_1 and CONS_1 appear in the workfile Return to Proc Make Graph in the model win dow choose Listed Vari ables list Y and check Compare As you can see in the dialog many options are available We re asking for a comparison of the baseline solution to the new sce nario We can show the dif ference between the two by checking one of the Devi ations boxes in either units or as a percentage Hint You can override an exogenous variable but you cannot override an endogenous variable because the latter would require a change to the structure of the model EViews Illustrated book Page 377 Monday February 25 2013 10 06 AM 378 Chapter 15 Super
196. ed book Page 11 Monday February 25 2013 10 06 AM 12 Chapter 1 A Quick Walk Through To get started we need to create the variable t In the command pane at the top of the EViews screen type series t trend TREND is one of the many functions built into EViews for manipulating data Double click on and you ll see something like the screen shown Since we want to think about how volume behaves over time we want to look at the variables T and LOGVOL together In EViews a collection of series dealt with together is called a Group To create a group including T and LOGVOL first click on Now while holding down the Ctrl key click on Then right click highlighting Open bringing up the context menu as shown and choose as Group The group shows time and log volume that is the series T and LOGVOL together Just as there are multiple ways to view a series there are also a number of group views Here s the spreadsheet view EViews Illustrated book Page 12 Monday February 25 2013 10 06 AM Estimating your first regression in EViews 13 Looking at a spreadsheet of a group with two series leaves us in the same situation we were in earlier with a spreadsheet view of a single series too many numbers A good way to look for a rela tionship between two series is the scatter dia gram Click on the button and choose Graph Then select Scatter as the Graph Type on the left hand side of the dialog that pops up
197. edit mode off gives you a little protection against making an accidental change and economizes on screen space by suppressing the edit field in the spreadsheet This is purely a matter of personal taste 105 106 EViews Illustrated book Page 409 Monday February 25 2013 10 06 AM 410 Chapter 18 Optional Ending Workfile Storage Defaults Back in the old days computer storage was a scarce commodity Data was often stored in single precision offering about seven digits of accuracy in four bytes of storage Today data are usually represented in double precision giving 16 digits of accuracy in eight bytes of storage Since raw data aren t likely to be accurate to more than seven digits single precision seems sufficient However numerical operations can introduce small errors EViews holds all internal results in double precision for this reason Using double precision when storing the workfile on the disk preserves this extra accuracy Using single precision cuts file size in half but causes some accuracy to be lost Internally each observation in a series takes up eight bytes of storage There s no great rea son you should care about this as either you have enough memory in which case it doesn t matter or you don t have enough memory in which case your only option is to buy some more And as a practical matter unless you re using millions of data points the issue will never arise Use compressi
198. ee the User s Guide or heck just click on the relevant menu and see what you get EViews Illustrated book Page 207 Monday February 25 2013 10 06 AM 208 Chapter 7 Look At Your Data Boxplots Sometimes a pic ture is better than a table Boxplots also called box and whisker dia grams pack a lot of information about the distri bution of a series into a small space The vari ety of options are controlled in the Graph Ele ments Boxplots page of the Graph Options dialog A boxplot of GPA using EViews defaults is shown here Opening the boxplot The top and bottom of the box mark 75th and 25th percentile of the distribution The dis tance between the two is called the interquartile range or IQR because the 75th percentile marks the top quartile the upper fourth of the data and the 25th percentile marks the bot tom quartile the bottom fourth of the data EViews Illustrated book Page 208 Monday February 25 2013 10 06 AM Describing Series Picturing the Distribution 209 The width of the box can be set to mean nothing at all the default or to be proportional to the number of observations or the square root of the number of observations Use the Box width radio buttons in the dialog The mean of the data is marked with a solid round dot The median of the data is marked with a solid horizontal line Shading around the horizontal line is used to compare differ ences between medians overlapping shade
199. elevant However which uncon ditional tables are shown does depend on the order However the series order does affect readability It generally makes sense to put first the variables you re most interested in com paring These will be the ones that show up together on each table Another approach to improved readability is to arrange series for the easiest screen display The two rules are The second series should have sufficiently few categories such that the categories can go across the top of the table without forcing you to scroll horizontally If you re going to print think about the width of your paper instead of the width of the screen EViews Illustrated book Page 220 Monday February 25 2013 10 06 AM Describing Groups Just the Facts Putting It Together 221 The first series should have as many categories as possible It s easier to look at one long table rather than many short tables This rule is sometimes limited by the desire to get a complete table to fit vertically on a screen or printed page True Story to End the Chapter When the author was a college student he worked as a research assistant for a professor from whom he learned a great deal about many things One incident was particularly mem orable We had turned in a report including relevant cross tabulations to the government agency which had paid for the research project Shortly thereafter a somewhat snippy letter came back pointing out
200. en panels and pools We look at an example and then discuss some of the nuances that help choose which is the better setup for a particular application Pooled or Paneled Population We just happen to have annual data on U S and Canadian population The workfile Pop_Pool_Panel wf1 contains a page named Pool with the pooled population and a page named Panel with the paneled population EViews Illustrated book Page 269 Monday February 25 2013 10 06 AM 270 Chapter 10 Prelude to Panel and Pool The picture on the left shows the pool approach which is pretty straightforward The data run for 51 years stretching from 1950 through 2000 The two series POPCAN and POPUSA hold values for the Canadian and U S population respectively The object ISOCODE is called a pool ISOCODE holds the words CAN and USA to tell EViews that POPCAN and POPUSA are series measuring POP for the respective countries In Chapter 12 Everyone Into the Pool we meet a variety of features accessed through the pool object that let you process POPCAN and POPUSA either jointly or separately But if you didn t care about the pool aspect you could treat the data as an ordinary EViews workfile So one advantage of pools is that the learning curve is very low The picture on the right shows the panel approach which introduces a kind of structure in the workfile that we haven t seen before The Range field now reads 1950 2000 x 2 The da
201. ence bands for example after forecasting a series As an example we estimated log GDP as a function of unem ployment and a time trend and then used EViews forecasting fea ture to put the forecast values of GDP in the series GDPC96F and the forecast standard errors in GDPC96SE The command show gdpc96f 1 96 gdpc96se gdpc96f 1 96 gdpc96se gdpc96f Hint Use an error bar graph whenever you want to draw primary attention to a central point the third series and secondary attention to a range EViews Illustrated book Page 145 Monday February 25 2013 10 06 AM 146 Chapter 5 Picture This opened a group window which we then switched to an Error Bar graph We added the title manually Of course EViews can also pro duce forecast graphs with confi dence intervals automatically See Chapter 8 Forecasting Scatter Plots Scatter plots are used for looking at the relation between two or more variables We ll use data on undergraduate grades and LSAT scores for applicants to the University of Washing ton law school to illustrate scatter plots Law 98 wf1 Simple Scatter At the default settings Scatter creates a scatter plot using the first series in the group for X axis values and the second series for the Y axis While there s a ten dency for higher undergraduate grades to be associated with higher LSAT scores as shown on the graph to the right the rela tionship certainly isn t a v
202. ent Each series in a workfile begins at the first date in the workfile and ends at the last date in the workfile What s more the existence of one series isn t related to the existence of some other series That is the series may be related by economics and statistics but EViews sees them as objects that just happen to be collected together in one place To pick a prospicient example we might have one annual data series on U S population and another on Canadian population EViews doesn t understand that the two series contain related observations on a single variable population But it might be convenient if EViews did understand no Not only does EViews have a way to tie together these sort of related series EViews has two ways panels and pools Panels are discussed in depth in Chapter 11 Panel What s My Line and we cover pools in Chapter 12 Everyone Into the Pool Here we do a quick compare and contrast A panel can be thought of as a set of cross sections countries people etc where each place or person can be followed over time Panels are widely used in econometrics A pool is a set of time series on a single variable observed for a number of places or people Pools are very simple to use in EViews because all you need to do is be sure that series names follow a consistent pattern that tells EViews how to connect them with one another In other words there s a great deal of overlap betwe
203. eralized Method of Moments 347 Thus to get 2SLS results we can give the command tsls inf c inf 1 unrate 1 c unrate 1 inf 1 inf 2 The coefficient on future inflation is now close to 1 0 as theory pre dicts The coefficient on unemploy ment is negative albeit small and not significant Did you notice the in the 2SLS output It s negative This means that the equation fits the data really poorly That s okay Our interest here is in accurate parameter estimation Generalized Method of Moments What happens if you put together nonlinear estimation and two stage least squares While EViews will happily estimate a nonlinear equation using the tsls command nowadays econometricians are more likely to use the Generalized Method of Moments or GMM Two stage least squares can be thought of as a special case of GMM GMM extends 2SLS in two dimensions GMM estimation typically accounts for heteroskedasticity and or serial correlation GMM specification is based on an orthogonality condition between a possibly nonlin ear function and instruments As an example suppose instead of the tsls command above we gave the gmm command gmm inf c inf 1 unrate 1 c unrate 1 inf 1 inf 2 Hint By default if you don t include the constant C in the instrument list EViews puts one in for you You can tell EViews not to add the constant by unchecking the Include a constant box in the estimation dialog
204. eries names 297 312 apostrophe 84 quotes in strings 108 plus function 105 equal function 84 87 105 A academic salaries example 23 24 across plot 158 add factors 380 alert boxes customizing 404 aliases 369 alpha series 102 103 alpha truncation 404 409 annualize function 108 109 apostrophe 84 ARCH autoregressive conditional heteroskedastic ity 351 355 ARCH in mean ARCH M 354 area graphs 132 arguments program 382 383 ARIMA AR Integrated MA model 329 333 arithmetic operators 86 87 arithmetic computer 100 ARMA errors 229 323 324 329 ARMA model 329 333 Asian Americans as group 286 aspect ratio 183 audit trail creating 382 autocorrelations 318 319 autoregressive conditional heteroskedasticity ARCH 351 355 autoregressive moving average ARMA errors 229 323 324 329 autoregressive moving average ARMA model 329 333 auto series 92 95 103 averages geometric 266 267 axes scales 141 axes customizing graph 140 141 148 184 186 axis borders 136 axis labels 163 B backing up data 243 399 bang symbol 384 bar graph 130 bar graphs 133 baseline model data 375 376 378 binning See grouping BMP 126 box plots 208 210 214 Box Jenkins analysis 329 333 Breusch Godfrey statistic 320 321 C C keyword 14 canceling computations 401 capitalization 28 85 279 caret symbol 228 case sensitivity 28 85 279 categorical gr
205. ers in Washington State in the workfile CPSMAR2004WA wf1 The series UNION is coded as one for union members and zero for non members Between four and five percent of workers in our sample are members of a union Is age an important determinant of union membership We might run a regression to see According to the least squares results age is highly significant statistically the t statistic is 2 8 but doesn t explain much of the variation in the dependent vari able the is low We can also look at the regression on a scatter plot The dependent variable is all zeros and ones The predicted values from the regres sion lie on a continuous line While the regression results aren t necessarily wrong what does it mean to say that predicted union membership is 0 045 Either you are a member of a union or you are not a member of a union This example is a member of a class called limited dependent R2 EViews Illustrated book Page 349 Monday February 25 2013 10 06 AM 350 Chapter 14 A Taste of Advanced Estimation variable problems EViews provides estimation methods for binary dependent variables as in our union membership example ordered choice models censored and truncated models tobit being an example and count models The User s Guide provides its usual clear expla nation of how to use these models in EViews as well as a guide to the underlying theory We ll illustrate with the simplest m
206. ery strong one If there were a really strong relationship law schools wouldn t need to look at both grades and test scores Hint Scatter plots don t give good visuals when there are too many observations The plots tend to get too busy to discern patterns Additionally scatter plots with very large numbers of observations can take a very very long time to render when the graphs are copied outside of EViews EViews Illustrated book Page 146 Monday February 25 2013 10 06 AM Group Graphics 147 Scatter with Regression EViews offers several options for fitting a line or curve to the data in a scatter plot in the Fit Lines menu Regression Line is the one most commonly used To learn about the other options see the User s Guide As you might expect Regres sion Line adds a least squares regression line to the scatter plot we ve just seen If you want to fit the line to transformed data using logs for example choose to bring up the Scatterplot Cus tomize dialog to choose from a variety of transformations Hint The equation for the line shown is of course the equation estimated by the com mand ls lsat c gpa EViews Illustrated book Page 147 Monday February 25 2013 10 06 AM 148 Chapter 5 Picture This Multiple Scatters If the group has more than two series EViews presents a number of choices in the Multiple series field of the Graph Options dialog The default is to add more scatterin
207. es Backwards compatibility hint The page feature was first introduced in EViews 5 There are three options if you want a friend with an earlier version to be able to read the data Tell them to upgrade to the current release There are lots of nifty new features Early versions of EViews will read the first page of a multi page workfile just fine but ignore all other pages So if the workfile has only a single page or if the page of interest is the first one created the left most page on the row of page tabs then there s no problem If not you can reorder the pages by dragging the page tabs and dropping them in the desired position Keep in mind that when reading a file earlier versions ignore object types that hadn t yet been invented when the earlier version was released Use Save Workfile Page to save the page of interest in a standalone workfile EViews Illustrated book Page 243 Monday February 25 2013 10 06 AM 244 Chapter 9 Page After Page After Page Multi Page Workfiles The Most Basic Motivation We ll get to some fancy uses of pages shortly But don t overlook the simplest reason for using multi page workfiles If you have sets of data that you want to keep in one collection just make each one a page in a workfile As in the example appearing to the right it s perfectly okay if the sets of data are unrelated To use a particular page click on the appropriate tab at the bottom of the
208. es EViews Illustrated book Page 115 Monday February 25 2013 10 06 AM 116 Chapter 4 Data The Transformational Experience EViews Illustrated book Page 116 Monday February 25 2013 10 06 AM Chapter 5 Picture This Interest rates over a wide spectrum of maturities three months to 20 years mostly move up and down together Long term interest rates are usually although not always higher than short term interest rates Long term interest rates also bounce around less than short term interest rates One picture illustrates all this at a glance This chapter introduces EViews graphics EViews can produce a wide variety of graphs and making a good looking graph is trivial EViews also offers a sophisticated set of customiza tion options so making a great looking graph isn t too hard either In this chapter we focus on the kinds of graphs you can make leaving most of the discussion of custom settings to Chapter 6 Intimacy With Graphic Objects We start with a simple soup to nuts example showing how we created the interest rate illustration above This is followed by simpler examples illustrating first graph types for single series and next graph types for groups of series A Simple Soup To Nuts Graphing Example The workfile Treasury_Interest_Rates wf1 contains monthly observations on interest rates with maturities from three months to 20 years Multiple series are plotted together in the same way that EViews a
209. es other than Raw data produce a single summary statistic for each group of data If all you have is one series there s only one number to plot Plotting sum mary statistics gets interesting when you compare statistics for different groups of data We saw this in the comparison of the means of three different interest rate series in the plot of the average yield curve on page 130 We ll see examples where the groups of data represent different categories in the next section Hint Rotated only works for some graph types For types where it doesn t the Rotated option won t appear Hint Frozen graphs with updating off don t rotate Continuing hint But if you wish you can accomplish the same thing by going to the Axes amp Scaling section and reassigning the series manually EViews Illustrated book Page 156 Monday February 25 2013 10 06 AM Categorical Graphs 157 Categorical Graphs So far all our graphs have produced one plot per series EViews can also display plots of series broken down by one or more categories This is a great tool for getting an idea of how one variable affects another Categorical graphs work for both raw data and summary statistics and pretty much all the graph types available under Basic graph are also available under Categorical graph For example a bar graph of median wages isn t nearly so interesting as is a graph compar ing wages for union and non union workers Union workers get paid
210. es several different ways to look at the residuals Usually the best view to look at first is Actual Fitted Resid ual Actual Fitted Residual Graph as illustrated by the graph shown here Three series are displayed The residuals are plotted against the left vertical axis and both the actual log volume and fitted predicted log volume series are plotted against the vertical axis on the right As it happens because our fit is quite good and because we have so many observations the fitted values nearly cover up the actual val ues on the graph But from the residuals it s easy to see two facts our model fits better in the later part of the sample than in the earlier years the residuals become smaller in abso lute value and there are a very small number of data points for which the fit is really terri ble EViews Illustrated book Page 79 Monday February 25 2013 10 06 AM 80 Chapter 3 Getting the Most from Least Squares Points with really big positive or negative residuals are called outliers In the plot to the right we see a small number of spikes which are much much larger than the typical residual We can get a close up on the residuals by choosing Actual Fitted Residual Residual Graph It might be interesting to look more carefully at specific num bers Choose Actual Fitted Residual Actual Fitted Resid ual Table for a look that includes numerical values You can see enormous residuals in
211. es variation within a country averaged across the pooled sample This standard devia tion is just over 4 10ths of one percent So the typical country in our pooled sample has only slightly higher variability in population growth than the United States The choice to remove means or not before computing descriptive statistics isn t a right or wrong issue It s a way of answering different questions Cross Section Specific Statistics Choosing the Cross section specific radio button generates descriptive statistics for each country separately one column for each country for each series In our example we have six reports from D LOG POP six from Y and one from D LOG POPUSA an excerpt of which is shown below Empirical aside If you re following along on the computer you can scroll the output to see that three of the countries in the pooled sample have population growth stan dard deviations much lower than the U S and three have standard deviations a little above that of the U S EViews Illustrated book Page 300 Monday February 25 2013 10 06 AM Playing in the Pool Data 301 Time Period Specific Statistics Time period specific is the flip side of Cross section spe cific Time period specific pools the whole sample together and then computes mean median etc for each date You can save the time period specific statistics into series Click the button and choose Make Periods Stats series Check boxes fo
212. ession command EViews provides estimates for all three coefficients and Has adding to the equation given a better fit Let s take a look at the residuals for this new equation Click View Actual Fitted Resid ual Actual Fitted Residual Graph again The fitted line now does a much nicer job of matching the long run characteristics of In particular the residuals over the last several decades are now flat rather than trending strongly upward This new equation is noticeably better at fit ting recent data t2 a b 1 b 2 t2 volume log EViews Illustrated book Page 17 Monday February 25 2013 10 06 AM 18 Chapter 1 A Quick Walk Through Saving your work Quite satisfactory but this is getting to be thirsty work Before we take a break let s save our equation in the workfile Hit the button on the equation window In the upper field type a meaningful name Prior to this step the title bar of the equation window read Equation untitled Using the button changed two things the equation now has a name which appears in the title bar and more importantly the equation object is stored in the workfile You can see these changes below If you like close the equation window and then double click on to re open the equation But don t take the break quite yet Hint Spaces aren t allowed when naming an object in EViews EViews Illustrated book Page 18 Monday February 25 2013 10 0
213. etimes a picture is better than a number Open a series or group of series and choose the Graph view see Chapter 5 Pic ture This While all graphs look at data the Distribution Quan tile Quantile and Boxplot options bear directly on understanding how a set of data is distributed The Distribution option offers a whole set of options the most familiar one being the Histogram Histograms A histogram is a graphical repre sentation of the distribution of a sample of data For the GPAs we see lots of applications around 3 4 or 3 5 and very few around 2 0 EViews sets up bins between the lowest and highest observation and then counts the number of observations falling into each bin The number of bins is chosen in order to make an attractive pic ture Clicking the button leads to the Distribution Plot Customize dialog where several customization options are provided See the User s Guide EViews Illustrated book Page 203 Monday February 25 2013 10 06 AM 204 Chapter 7 Look At Your Data All the features described in Chapter 5 Picture This and in Chapter 6 Inti macy With Graphic Objects can be used for playing with histo grams For exam ple we can make a categorical graph to compare GPAs of Washington State residents to those of non resi dents EViews Illustrated book Page 204 Monday February 25 2013 10 06 AM Describing Series Picturing the Distribution 205 Cautionar
214. ets the left facing bar These graphs carry a lot of infor mation They re probably most effective when limited to a small number of data points The ver sion shown to the right covers two weeks whereas the previous graph had four months of data This shorter graph does a better job of showing market behavior in the period right around the 1987 stock market crash Orderly hint The order in which the series appear in the group matters for the High Low graph EViews puts series in the same order as you select them in the workfile window but the order is easily re arranged by choosing the Group Members view and manually editing the list This is true for any group but it s especially important for graphs such as the High Low graph where the order of the series affects their inter pretation Be sure to click the button when you re done re arranging EViews Illustrated book Page 144 Monday February 25 2013 10 06 AM Group Graphics 145 Error Bar Graphs Error bar graphs are similar to High Low graphs in that pairs of observations from the first two series are used to mark the high and low values of vertical lines They re slightly differ ent in that error bar graphs have small horizontal caps drawn at the top and bottom of each line When there s a third series its value is marked by placing a symbol on the vertical line Such triplet error bar graphs are commonly used for showing point estimates together with confid
215. ever method is set as the default in the series being copied Constant match average In the case we re looking at the conversion method applied by default was Constant match average This instruction parses into two parts constant and match average Constant means uses the same number in each month in a quarter Match average instructs EViews that the average monthly value chosen has to match the quarterly value for the corresponding quarter At this point you may be muttering under your breath What do these guys mean the average monthly value There only is one monthly value You have a good point When converting from low to high frequency saying constant match average is just a convo luted way of saying copy the value As we ve seen that s exactly what happened above Hint Low to high frequency conversion means from annual to quarterly quarterly to monthly annual to daily etc EViews Illustrated book Page 248 Monday February 25 2013 10 06 AM Multiple Frequencies Multiple Pages 249 Later we ll see conversion situations where constant match average is much more com plex Constant match sum Suppose that instead of GDP measured at an annual rate our low frequency variable hap pened to be total quarterly sales In a conversion to monthly sales we d like the converted January February and March to add up to the total of first quarter sales Constant match
216. everse relation is sometimes thought to signal that a recession is likely Here s a graph that shows the difference between the one year and three month inter est rates EViews Illustrated book Page 172 Monday February 25 2013 10 06 AM Templates for Success 173 Use to add a horizontal line at zero as shown In the revised figure it s much easier to see how rare it is for the longer term rate to dip below the short rate Templates for Success What s really useful is to see how the difference between long and short term interest rates compares to shaded periods of recession Periods in which the long rate dips below the short rate are associated with reces sions On the other hand there have been a number of reces sions in which this didn t hap pen In fact adding shading for reces sions has become something of a EViews Illustrated book Page 173 Monday February 25 2013 10 06 AM 174 Chapter 6 Intimacy With Graphic Objects standard in the display of U S macro time series data Including shades for the NBER reces sion dates is so easy that you may want to make it a regular practice Just use the button for each of the 32 recessions identified by the NBER You don t really want to re enter the same 32 sets of shade dates every time you draw a graph do you Templates to the rescue A template is nothing more than a graphic object from which you can copy option set tings into another g
217. eviews com or on a CD bundled with EViews Illustrated Despite all care an error or two undoubtedly remain Corrections comments compliments and caritas all gratefully received at EViewsIllustrated eviews com Dick Startz Castor Professor of Economics University of Washington Seattle February 2012 Hint Remember that this is a tutorial not a reference manual More is not necessarily better EViews comes with over 2 500 pages of first class reference material in four vol umes When details are better explained by saying See the User s Guide that s what we do And if 2 500 pages just isn t enough or is perhaps too much you can always visit the EViews Forum http forums eviews com where you can find answers to commonly asked questions share information and mingle with like minded EViews users EViews Illustrated book Page 1 Monday February 25 2013 10 06 AM 2 Foreword EViews Illustrated book Page 2 Monday February 25 2013 10 06 AM Chapter 1 A Quick Walk Through You and I are going to start our conversation by taking a quick walk through some of EViews most used features To have a concrete example to work through we re going to take a look at the volume of trade on the New York Stock Exchange We ll view the data as a set of numbers on a spreadsheet and as a graph over time We ll look at summary statis tics such as mean and median together with a histogram Then we ll build a simple regres
218. ew which lets us see the number for VOLUME at each date Another way to think about data is by look ing at summary statistics In fact some kind of summary statistic is pretty much necessary in this kind of situation we can t hope to learn much from staring at raw data when we have 400 plus numbers To look at summary statistics for a series press the button and choose Descriptive Statistics amp Tests Histogram and Stats Here we see a histogram which describes how often different values of VOLUME occurred Most periods trading is light Heavy trading volume say over a billion shares happened much less frequently To the right of the histogram we see a variety of summary statistics for the data The average mean volume was just over 93 million shares while the median volume was 1 75 million The largest recorded trading volume was over 1 6 billion and the smallest was just over 100 000 Now we ve looked at two different views Spreadsheet and Histogram and Stats of a series VOLUME We ve learned that trading volume on the NYSE is enormously variable Can we say something more systematic explaining when trading volume is likely to be high versus low A starting theory for building a model of trading volume is that trading volume grows over time Underlying this idea of growth over time is some sense that the financial sector of the economy is far larger than it was in the past EViews Illustrated book Page 5 Monday
219. ews Illustrated book Page 8 Monday February 25 2013 10 06 AM Generating a new series 9 Double click and then scroll the window so that the beginning of 2000 is at the top You ll see a win dow looking something like the one shown here Starting in 2001 we see numbers Before 2001 only the letters NA There are two lessons here EViews operates only on data in the current sam ple When we created LOGVOL the sample began in 2001 No values were generated for earlier dates EViews marks data that are not available with the symbol NA Since we didn t generate any values for the early years for LOGVOL there aren t any values available Hence the NA symbols Since we re trying to look at all the available data we want to change the sample to include well the whole sample One way to do this is to use the menu selection Quick Sample Another way to change the sample is to double click on the sample line in the upper pane of the workfile window right where the arrow s pointing in the picture on the right To illustrate another alternative we ll type our first command The workfile window and the series window appear in the lower section of the master EViews window The upper area is reserved for typing commands and not surprisingly is called the command pane The command smpl is used to set the sample the keyword all signals EViews to use all available data in the current sample Type the co
220. example we ve been using we ve taken our task to be forecasting currency growth If you look at the label for our series G see Label View in Chapter 2 EViews Meet Data you ll see it was derived from an underlying series for the level of currency CURR using the com mand series g 1200 dlog curr Since the function dlog takes first differences of logarithms we actually made both of the transformations just mentioned Instead of forecasting growth rates we could have been asked to forecast the level of cur rency In principle if you know today s currency level and have a forecast growth rate you can forecast next period s level by adding projected growth to today s level In practice doing this can be a little hairy because for more complicated functions it s not so easy to work backward from the estimated function to the original variable and because forecast confidence intervals see below are nonlinear Fortunately EViews will handle all the hard work if you ll cooperate in one small way Estimate the forecasting equation using an auto series on the left in place of a regular series As a regular series the information that G was created from g 1200 dlog curr is a his torical note but there isn t any live connection If we use an auto series then EViews understands and can work with the connection between CURR and the auto series We could use the following commands to define an a
221. f population EViews Illustrated book Page 279 Monday February 25 2013 10 06 AM 280 Chapter 11 Panel What s My Line Use the button to choose Descriptive Statistics amp Tests Stats by Classification Use ISOCODE as the classifying variable The average value of population growth was 2 2 percent per year in the Central African Republic and 1 6 percent in Canada If we multiply the difference in population growth rates 0 008 by the estimated regression coefficient 901 we predict that relative GDP in the Central African Repub lic should be 7 percentage points lower than in Canada Population growth appears to have a very large effect Hint Lags in panel workfiles work correctly in other words EViews knows that a lag means the previous obser vation for the same country Notice in the window to the right that the 1950 value of D LOG POP is correctly NA Even though the observation for the year 2000 for Can ada appears immediately before 1950 Switzerland in the spreadsheet EViews understands that the observations are not sequential Convenience hint It wasn t necessary to restrict the sample to the two countries of interest Limiting the sample just made the output window shorter and easier to look at EViews Illustrated book Page 280 Monday February 25 2013 10 06 AM Panel Estimation 281 Is the apparent effect of population growth on output real or is it a spurious result It s easy to imagi
222. f we make UNION FE the within factors FE codes gender we get the plot shown below Junk graphics alert The graph appears to show that union workers are paid enor mously more than are non union workers In generating a visually appealing graph EViews selected a lower limit of 24 000 for the vertical axis Union wages are about 60 percent higher than non union wages But the union bar is about 10 times as large The visual impression is very misleading We ll fix this in Chapter 6 page 185 EViews Illustrated book Page 158 Monday February 25 2013 10 06 AM Categorical Graphs 159 Note that the first categorical variable listed becomes the major grouping control Switch the within order to FE UNION and EViews switches the ordering in the graph as shown below Hint The graphs give identical information but the first graph gives visual emphasis to the fact that men are paid more than women whether they re union or non union The second graph emphasized the union wage premium for both genders EViews Illustrated book Page 159 Monday February 25 2013 10 06 AM 160 Chapter 5 Picture This If we move the variable FE to the Across field EViews splits the graph across separate plots for each gender You can use as many within and across factors as you wish We ve added highest degree received to the across factor in this graph EViews Illustrated book Page 160 Monday February 25 2013 10 06 AM Categori
223. fect of growth is much smaller The standard error is also smaller but it shrunk by less than the coef ficient did It would have been nicer if the t statistic had become larger rather than smaller at least if nicer is interpreted as provid ing support to our preconceived notions As you might guess the menu choice Period weights is analo gous to Cross section weights allowing for different variances for each time period instead of each country EViews Illustrated book Page 307 Monday February 25 2013 10 06 AM 308 Chapter 12 Everyone Into the Pool Cross country Correlations To account for correlation of errors across countries as well as different variances choose Cross section SUR This option requires balanced samples ones that all have the same start and end dates So if your pool isn t balanced ours isn t you ll also need to check the Balance Sample check box at the lower right of the dia log The estimated effect of population growth is now quite small but statistically very signifi cant Period SUR provides the analogous model where errors are correlated across periods within each country s observations More Options to Mention If you re in an exploring mood note that EViews will do random effects as well as fixed effects in the Estimation method field of the Specification tab of the Pool Estimation dia log and Period specific coefficients just as it does Cross section spe
224. ficient values at once as in param c 1 3 14159 c 2 2 718281828 volume atb R2 volume a 1 t b volume atb EViews Illustrated book Page 344 Monday February 25 2013 10 06 AM 2SLS 345 Change iteration limits If the estimate runs but doesn t converge give the same command again Since EViews stores the last estimated coefficients in C the second estimation run picks up exactly where the first one left off Alternatively click the but ton and switch to the Options tab Try increasing Max iterations You can also put a larger number in the Convergence field if you re will ing to accept potentially less accu rate answers 2SLS For consistent parameter estima tion the sine non qua assumption of least squares is that the error terms are uncorrelated with the right hand side variables When this assumption fails econometricians turn to two stage least squares or 2SLS a member of the instrumental vari able or IV family 2SLS augments the information in the equation specification with a list of instruments series that the econometrician believes to be correlated with the right hand side variables and uncorrelated with the error term As an example consider estimation of the new Keynesian Phillips curve in which infla tion depends on expected future inflation and unemployment possibly lagged In the fol lowing oversimplified specification pt a bpe t 1 gunt 1
225. for command with a control variable is for control_variable first_value to last_value step stepvalue some commands go here next If the step value is omitted EViews steps the control variable by 1 The program count prg counts to 100 displaying the count on the status line Most econometricians can count to 100 on their own so this program is rarely seen in the wild We captured this unusual specimen because it provides a particularly pristine example of a numerical loop For something a bit more likely to be of practical use we ve simplified silly_program prg by putting the values 0 and 1 in a loop instead of writing out each statement twice In less_silly_program prg GEN DER is the control variable FIRST_VALUE is 0 and LAST_VALUE is 1 String Loops A for command with a string variable has a string variable followed by a list of strings no equal sign EViews Illustrated book Page 385 Monday February 25 2013 10 06 AM 386 Chapter 16 Get With the Program Each string is placed in turn in the string variable and the lines between for and next are executed for string string1 string2 string3 some commands go here next The commands between for and next are executed first with the string string1 replacing string then with string2 etc To further simplify less_silly_program prg so that we don t have to write our code separately for ED and for AGE we
226. for each series separately The Current Population Sur vey isn t limited to workers so the state by state means have been computed as a fraction of the population We probably wanted only those who are working What s more the vari able LNWAGE is coded as NA for anyone who doesn t report a positive salary including all non workers As a result the state by state means for LNWAGE were computed using roughly 25 percent fewer observations than the other variables We want a common sample to be used for computing the series means for each state This can be accomplished by specifying an appropriate sample in the Source Sample field in the Paste Special dialog It happens that in this data set the only difference in the sample for the different series is that LNWAGE has a lot of NAs Toss out the ByState page we made and make a new one this time entering if not isna lnwage in the Source Sample field Hint We could have edited the link specifications but since we had several series it was faster to just toss the links and start over EViews Illustrated book Page 263 Monday February 25 2013 10 06 AM 264 Chapter 9 Page After Page After Page We now have a valid state by state dataset Let s repeat our earlier regression using state level data The results are basi cally the same The estimated effects of both education and age are a little larger than for the individual data While the coeffi cients are
227. from many popular statis tics packages EViews does an excellent job of reading these formats while preserving label ing information So if you are given data from Stata TSP SPSS SAS etc try reading the data directly using File Open Foreign Data as Workfile See the User s Guide for more information about this too Hint Telling EViews to use fixed column locations for each series replaces the use of delimiters This can get around the missing number problem described above Simi larly it can solve the problem of alpha observations that include spaces that would be mistaken for delimiters EViews Illustrated book Page 47 Monday February 25 2013 10 06 AM 48 Chapter 2 EViews Meet Data Coming in from the clipboard Another useful option is to use Copy Paste For example some web sites load data right onto the clipboard EViews is happy to create a new work file from the contents of the clipboard Right click on any blank spot in the lower EViews pane and then choose Paste as new Workfile EViews will try to use the top row of data for series names If that doesn t work out EViews will name the series SERIES01 SERIES02 etc in which case you may want to rename the series to something more meaningful To rename a series select the series in the work file window right click and choose Rename Fill out the dialog with the new name You can enter a Display Name here as well Reading From the Web EView
228. ge 312 Monday February 25 2013 10 06 AM Quick Review Pools 313 coefficients for all countries You can also allow for individual coefficients by cross section or by period for any variable Fixed and random effect estimators are built in And because pooled series are just plain old series with a clever naming convention all of the EViews fea tures are directly available See Chapter 11 Panel What s My Line for a different approach to two dimensional data EViews Illustrated book Page 313 Monday February 25 2013 10 06 AM 314 Chapter 12 Everyone Into the Pool EViews Illustrated book Page 314 Monday February 25 2013 10 06 AM Chapter 13 Serial Correlation Friend or Foe In a first introduction to regression it s usually assumed that the error terms in the regres sion equation are uncorrelated with one another But when data are ordered for example when sequential observations represent Monday Tuesday and Wednesday then we won t be very surprised if neighboring error terms turn out to be correlated This phenomenon is called serial correlation The simplest model of serial correlation called first order serial cor relation can be written as The error term for observation t carries over part of the error from the previous period and adds in a new innovation By convention the innovations are themselves uncorrelated over time The correlation comes through the term If then 90 percent
229. gmstcen droplast gives the results to the right The regression is iden tical to our earlier fixed effect results You do have to remember that the constant term has a different inter pretation In the fixed effect panel estima tion the reported con stant is the average and the reported fixed effects are the devia tions from that average for each category When using expand Econometric reminder The dummy variable trap is what catches you if you attempt to have an intercept and a complete set of dummies in a regression ai EViews Illustrated book Page 288 Monday February 25 2013 10 06 AM Quick Review Panel 289 the reported constant is the intercept for the dropped category and the reported dummy coefficients are the difference between the category intercept and the intercept for the dropped category Quick Review Panel The panel feature lets you analyze two dimensional data Convenient features include pret tier identification of your data in spreadsheet views and some extra graphic capabilities The use of fixed effects in regression is straightforward and often critical to getting meaningful estimates from regression by washing out unobservables The examples in this chapter used cross section fixed effects but you can use period fixed effects or both cross section and period fixed effects just as easily See Chapter 12 Everyone Into the Pool for a different approach to two dimensional data
230. gression specification We want for our depen dent variable and a time trend for our independent variable Fill out the equation dialog by entering log volume c trend Hint EViews tells one item in a list from another by looking for spaces between items For this reason spaces generally aren t allowed inside a single item If you type log volume c trend you ll get an error message volume log EViews Illustrated book Page 63 Monday February 25 2013 10 06 AM 64 Chapter 3 Getting the Most from Least Squares Our regression results appear below The Really Important Regression Results There are 25 pieces of information displayed for this very simple regression To sort out all the different goodies we ll start by showing a couple of ways that the main results might be presented in a scientific paper Then we ll discuss the remaining items one number at a time A favorite scientific convention for reporting the results of a single regression is display the estimated equation inline with standard errors placed below estimated coefficients looking something like Exception to the previous hint When a text string is called for in a command spaces are allowed inside paired quotes Reminder The letter C in a regression specification notifies EViews to estimate an intercept the parameter we called above Hint Another reminder trend is an EViews function to generate a time trend 0 1
231. gression line through the cloud of points plotted with on the vertical axis and time on the horizontal The regression line can be written as an algebraic expression volume log volumet log a bt EViews Illustrated book Page 61 Monday February 25 2013 10 06 AM 62 Chapter 3 Getting the Most from Least Squares Using EViews to estimate a regression lets us replace and with numbers based on the data in the workfile In a bit we ll see that EViews estimates the regression line to be In other words the intercept is estimated to be 2 6 and the slope is estimated to be 0 017 Most data points in the scatter plot fall either above or below the regression line For exam ple for observation 231 which happens to be the first quarter of 1938 the actual trading volume was far below the predicted regression line In other words the regression line contains errors which aren t accounted for in the esti mated equation It s standard to write a regression model to include a term to account for these errors Econometrics texts sometimes use the Greek letter epsilon rather than u for the error term A complete equation can be written as Regression is a statistical procedure As such regression analysis takes uncertainty into account Along with an estimated value for each parameter e g we get Measures of the accuracy of each of the estimated parameters and related information for computi
232. gs to the plot using the third series versus the first the fourth series versus the first and so on Alternatively choosing Multiple graphs First vs All puts each scatter in its own plot You can also see all the possible pairs of series in the group in individual scatter plots by choosing Scatterplot matrix It s clear that the 3 month rate is more closely related to the 1 year rate than it is to the 20 year rate Not surprising of course Hint Changing the axis scales to logarithmic is different than drawing the fitted line using logs in the Scatterplot Customize dialog The former changes the display of the scattered points while the latter changes how the line is drawn Changing the axis scales is covered in Chapter 6 Intimacy With Graphic Objects EViews Illustrated book Page 148 Monday February 25 2013 10 06 AM Group Graphics 149 Where Scatterplot matrix gives each series a turn on both horizon tal and vertical axes Lower trian gular matrix shows a single orientation for each pair If you have four or more series in a group Scatter adds the choice Multiple Graphs XY Pairs which plots a scatter of the second series on the vertical axis versus the first series on the horizontal axis the fourth series versus the third series and so forth Contrast with Multiple graphs First vs All which uses the first series for the horizontal axis and places all the remaining series on the vertical axis Togethe
233. h Save Date Representation If you re American which includes Cana dian for the very lim ited purpose of this sentence skip this paragraph Americans write dates Month Day Year Most of the world prefers the order Day Month Year If you operate in the latter area click the Day Month Year radio button EViews Illustrated book Page 405 Monday February 25 2013 10 06 AM 406 Chapter 18 Optional Ending Spreadsheet Defaults Americans can skip this one too Americans separate the integer and fractional parts of numbers with a deci mal point If you prefer a comma check Comma as decimal in the Spread sheets Data display dialog The Data displayed as drop down determines what is shown in the rest of the Numeric display group This allows you to set the numeric display characteristics individually for each type of data in the list To switch all your data to Comma as decimal you ll want to switch to each item in the Data displayed as drop down and select Comma as decimal You do not need to click OK and exit the dialog between each setting EViews remembers each group of settings individually until you click OK More Detailed Options My personal recommendation If you ve made the changes above don t worry about other option settings for now As you become more of a power user you will find some personal customization helpful The remainder of this chapter is devoted to customization hints EViews Illu
234. h the following two commands given in the command pane not in the system window coef 14 j coef 14 k EViews provides a long list of estimation methods which can be applied to a system Click the button to bring up the System Estimation dialog and then choose an esti mation method from the Method dropdown Hint It helped to copy and paste from the representations view of the earlier least squares results and then to use Edit Replace to change the coefficients from C to K and J EViews Illustrated book Page 358 Monday February 25 2013 10 06 AM System Estimation 359 Choosing Ordinary Least Squares produces estimates for both equations The output is long only part is shown here Note that the results for cur rency in the hands of the public are precisely the same as those we saw previously We asked for equation by equation ordi nary least squares and that s what we got the equivalent of a bunch of ls commands EViews Illustrated book Page 359 Monday February 25 2013 10 06 AM 360 Chapter 14 A Taste of Advanced Estimation Instead of equation by equa tion least squares we might try a true systems estimator such as seemingly unrelated regres sions SUR The upper portion of the SUR results is shown to the right The estimated coefficients haven t changed much in this example The difference between the two esti mates is that the latter accounts for correlation between the
235. hat if one equation is misspecified that misspecification will pollute the estimation of all the other equations in the system To create a system object either give the system com mand or use the menu Object New Object Enter one or more equation specifi cations in the text area We ve added data on the growth in bank vault cash GV to our data set on growth in currency in the hands of the public G The specifica tion shown is identical for both cash components Hint Unlike nearly all other EViews estimation procedures maximum likelihood won t deal with missing data The series defined by logl must be available for every observation in the sample Define an appropriate sample in the Estimation dialog If you accidentally include missing data EViews will give an error message identifying the offending observation Worthy of repetition hint If you want an estimated coefficient to have the same value in more than one equation system estimation is the only way to go Use the same coefficient name and number e g C 3 in each equation you specify The jargon for this is constraining the coefficients Note that you can constrain some coefficients across equations and not constrain others EViews Illustrated book Page 357 Monday February 25 2013 10 06 AM 358 Chapter 14 A Taste of Advanced Estimation Before we can estimate the system shown we need to create the coefficient vectors K and J That can be done wit
236. he group The tests assume that the series are statistically independent Here too you can be sure the same observations are used from each series by picking Common sample Hint Use Tests For Descriptive Statistics Simple Hypothesis Tests for each individ ual series to test for a specific value say that the mean equals 3 14159 Use Tests For Descriptive Statistics Equality Tests By Classification to test that different sub populations have the same mean for a series Use Tests of Equality for groups to see if the mean is equal for different series EViews Illustrated book Page 214 Monday February 25 2013 10 06 AM Describing Groups Just the Facts Putting It Together 215 Correlations It s easy to find the covari ances or correlations of all the series in a group by choosing the Covariance Analysis view By default EViews will display the covariance matrix for the common samples in the group but you may instead compute correlations use individual samples or com pute various other measures of the associations and related test statistics For example to compute using individual samples you should uncheck the Balanced sample listwise deletion box This setting instructs EViews to compute the covariance or correlation for the first two series using the observations avail able for both series then the correlation between series one and series three using the observations available for those two series in
237. hese don t work once a graph is frozen with updating turned off A Touch of Text Since adding text is the easiest touch up we ll start there Hint Single click on almost any thing in a graphics window and hit the Delete key to make the thing disappear Be careful there s no undo and no Are you sure alert Hint I find it convenient to freeze a graph before fine tuning but to leave the original series or group window open until I m sure I won t need to make any changes that don t work on completely frozen graphs Alternately you can freeze the graph with updating turned on and make desired changes If you d like you can then copy the result or turn updating off once you are certain that you are done Hint To make side by side comparisons of different visual representations of a given set of data you have to freeze at least one window because you can only have a single view open of a series or group Freezing creates a new graph object detached from the original view Hint As you know you must freeze a graph before you can add text to it EViews Illustrated book Page 168 Monday February 25 2013 10 06 AM A Touch of Text 169 Clicking opens the Text Labels dia log Enter the desired text pick a location and hit Note that the Enter key is intentionally not mapped to the but ton so that you can enter multi line labels by ending lines with Enter The options on the dialog mean pretty much jus
238. hether or not the individual is a union member UNION 1 if union member 0 if not The identifier of this data set is the observation num ber for a particular individual Our goal is to regress log wage on education age union membership and the fraction of the population that s unionized in the state The difficulty is that the unionized fraction of the state s population is naturally identified by state We need to find a mechanism to match individual identified data with the state identified data We ll do this in several steps EViews Illustrated book Page 258 Monday February 25 2013 10 06 AM Matching When The Identifiers Are Really Different 259 Let s first make sure that union ization matters at least for the person in the union The regres sion results here show a very strong union effect Controlling for education and age being a union member raises your wage by about 25 percent Specifying A Page By Identifier Series Our next step is to create a page holding data aggregated to the state level We want our new page to contain one observa tion for each of the states observed in the individual data Clicking on the New Page tab brings up a menu including the choice Specify by Identifier Series which not surpris ingly is just what we need to specify an identifier series for a new page In our original page the state identifier is in the series GMSTCEN so that s what we ll use as the identifier for the
239. holesale Link Creation We looked at the copy part of Copy Extract from Current Page in the section Creating New Pages We turn now to the By Link to New Page option By Link to New Page works just like Value to New Page or Workfile except that links are made for each series instead of copying the values into disconnected series in the desti nation page That s it One Or More At A Time Link Creation The most common way of creating a link is by copying series on one page and then pasting special on the destination page Inside the Paste Special dialog choose Link in the Paste as field Links for the series with frequency conversions as speci fied will be placed in the destination page Note that Copy Extract from Current Page always uses the default frequency conversion for each series Paste Special has the same default behavior but provides an opportunity in the Frequency conversion options field to make an idiosyncratic choice Inconsequential hint Links save computer memory because only one copy of the data is needed They use extra computer time because the linked data has to be regenerated each time it s used Modern computers have so much memory and are so fast that these issues are rarely of any consequence Hint You can link between pages inside a workfile but not across different workfiles EViews Illustrated book Page 252 Monday February 25 2013 10 06 AM Links The Live Connection 253 Retail Link
240. i e multiple series or multiple cate gories how would you like them visually arrayed Multiple graphs All in one graph Hint It s best to do as much graph editing as possible inside EViews before export ing so that EViews has a chance to touch up the final picture See Graphic Auto Tweaks above Hint There isn t any way to read a graphic file into EViews nor can you paste a pic ture from the clipboard into an EViews object EViews Illustrated book Page 126 Monday February 25 2013 10 06 AM A Graphic Description of the Creative Process 127 While it s helpful to think of these as four independent choices there is some interaction among them For example the number of series in the window determines the choices of graph types that are available A scatter diagram requires at least two series right The Graph Options dialog adjusts itself to display options sensible for the data at hand Thinking about the four basic choices in graph creation is a useful organizing principle but the truth is most graphs are made with a couple of mouse clicks and where to click is usu ally pretty obvious We ve been showing our interest rate data in a single graph as a useful way to show that interest rates of different maturities largely move together To show each series separately set the Multiple series dropdown menu to Multiple graphs Presto Stressing out Making a graph is starting to sound awfully complicated
241. ically or you can specify your preferred conversion method More on this below EViews Illustrated book Page 244 Monday February 25 2013 10 06 AM Multiple Frequencies Multiple Pages 245 production numbers are available more often and sooner they re also used to predict GDP Our task is to use industrial production to forecast GDP Initially our workfile has two pages Indpro has monthly industrial production data and Gdpc96 has quarterly GDP data To analyze the relation between RGDP and IP we have to get them into the same page for two reasons Mechanically only one page is active at a time so everything we want to use jointly had better be on that page To relate one series to another at least if we want to use a regression observations have to match up This implies that all the series in a regression need to have the same identifier and therefore the same fre quency EViews provides a whole toolkit of ways to move series between pages and to convert fre quencies which we ll take a look at now Hint http research stlouisfed org fred2 is an excellent source for U S macroeco nomic data Among other virtues FRED can make you an Excel file that EViews will read in about two seconds flat nother hint You can use EViews to search and read data directly from FRED Select File Open Database from the main EViews menu and select FRED database EViews will open a window for examining the FRED data
242. ically in the second Clinton administration and rebounded in the first Bush administration even while GDP growth was relatively steady High Low High Low Open Close Graphs High Low graphs take observa tions from pairs of series and draw vertical lines connecting each pair placing the value from the first series at the top and the value from the second series at the bottom The most common use of the High Low graph is to show opening and closing prices for a stock or other traded good but these graphs are nifty any time you want to display a range of values at each date The example here shows high and low prices for the Russell 3000 in the later part of 1987 Russell3000 wf1 October 19th really stands out doesn t it Hint The legend isn t displayed on High Low graphs so be sure to include your own label using the button as we have here EViews Illustrated book Page 143 Monday February 25 2013 10 06 AM 144 Chapter 5 Picture This High Low graphs display of pairs extends to display of triples or quadruples The first two series in the group mark the top and bottom of the vertical bar respectively If there are three series perhaps the third series represents closing prices values from the third series are shown with a right facing horizontal bar If there s a fourth series then it s shown with a right fac ing horizontal bar perhaps rep resenting closing prices and the third series g
243. ics to track the solution process are also avail able when needed Add factors can be used to adjust the value of a specified variable You can even use add factors to adjust the solution for a particular variable to match a desired target Equations can be implicit Given an equation such as EViews can solve for y Add factors can be used for implicit equations as well Stochastic simulations in which you specify the nature of the random error term for each equation are a built in feature This allows you to produce a statistical distribu tion of solutions in place of a point estimate Equations can contain future values of variables This means that EViews can solve dynamic perfect foresight models Single variable control problems of the following sort can be solved automatically You can specify a target path for one endogenous variable and then instruct EViews to change the value of one exogenous variable that you specify in order to make the solved for values of the endogenous variable match the target path Quick Review A model is a collection of equations either typed in directly or linked from objects in the workfile The central feature of the model object is the ability to find the simultaneous solu tion of the equations it contains Models also include a rich set of facilities for exploring var ious assumptions about the exogenous driving variables of the model and the effect of shocks to equations
244. ictures of the statistical distribution of the data rather than plotting the observations directly A histogram is prob ably the most familiar example These graphs are discussed in Chapter 7 Look At Your Data on page 195 Axis Borders Even though discussion of distribution graphs awaits Chapter 7 we ll sneak in one marginal comment EViews lets you decorate the axes of most graphs with small histograms or other dis tribution graphs by using the Axis borders menu This is a great technique for looking at raw data and distribution information together EViews Illustrated book Page 136 Monday February 25 2013 10 06 AM Group Graphics 137 The graph to the right has a his togram added to the line graph for the 3 month Treasury bill rate The histogram provides a reminder that interest rates were close to zero for much of our sample The line graph reveals that these extremely low rates were a phenomenon of the Great Depression and World War II Group Graphics Any graph type applicable to a single series can also be used to graph all the series in a Group EViews default setting is to plot the series in a single graph as in our interest rate example As you saw earlier you can switch the Multiple series field in Graph Options to Multiple graphs to get one series per graph EViews Illustrated book Page 137 Monday February 25 2013 10 06 AM 138 Chapter 5 Picture This Each of the graphs in the window
245. iduals are nearly all negative so the residuals are not centered on zero That s a hint that the country equations should have different intercepts We can get different intercepts by specifying fixed effects Let s look at the residual plots from the fixed effects equation we estimated This time each country s residual is centered around zero EViews Illustrated book Page 304 Monday February 25 2013 10 06 AM More Pool Estimation 305 Grabbing the Residual Series If six graphs in one window is looking a little hard to read think what sixty six graphs would look like Proc Make Residuals generates series for the residuals from each coun try RESIDCAN RESIDFRA etc and puts them into a group window From there it s easy to make any kind of group plot we d like Here s one to which we ve added a title and zero line It s pretty clear from this picture that the fit for Japan is problematic Our model doesn t take into account the post War Japanese recovery and it shows If this were a real research project we d have to stop and deal with the misspecification issue Employing the literary device suspension of disbelief we ll just proceed onward EViews Illustrated book Page 305 Monday February 25 2013 10 06 AM 306 Chapter 12 Everyone Into the Pool Residual Correlations Clicking on View Residu als Correlation Matrix gives us a table showing the correlation of the residuals The correl
246. iews Meet Data Frankly the easiest way to get data into EViews is to start with data that someone else has already entered into a computer file EViews is very clever about reading data in a variety of formats Let s go back to the academic salary example to go over some methods of bringing in data that s already on the computer EViews provides three different methods for loading data from a foreign file File Open Foreign Data as Workfile translates any of a number of file formats into an EViews workfile You can set up a workfile as we ve done above and then use File Import to bring in data from a spreadsheet program or text file plus a couple of other specialized for mats You can use the standard Windows copy paste commands to transfer data between a Series window or a Group window and another program EViews is quite smart about interpreting the material you re pasting If most of your data comes from a single source using File Open Foreign Data as Work file is far and away the easiest method If you re cobbling together data from multiple sources try using File Open Foreign Data as Workfile on the most complicated file and then using File Import or copy paste to add from the other sources one at a time EViews is a fluent reader of many foreign file formats Let s walk through examples of sev eral of the most common EViews Illustrated book Page 38 Monday February 25 2013 10 06 AM
247. iews lets you turn the process inside out by using the p value reported in the right most column under the heading Prob EViews has worked the problem backwards and figured out what size would give you a critical value that would just match the t statistic reported in column three So if you are interested in a five percent test you can reject if and only if the reported p value is less than 0 05 Since the p value is zero in our example we d reject the hypothesis of no trend at any size you d like Obviously that last sentence can t be literally true EViews only reports p values to four decimal places because no one ever cares about smaller probabilities The p value isn t liter ally 0 0000 but it s close enough for all practical purposes More Practical Advice On Reporting Results Now you know the principles of how to read EViews output in order to test whether a coef ficient equals zero Let s be less coy about common practice When the p value is under 0 05 econometricians say the variable is significant and when it s above 0 05 they say it s insignificant Sometimes a variable with a p value between 0 10 and 0 05 is said to be weakly significant and one with a p value less than 0 01 is strongly significant This practice may or may not be wise but wise or not it s what most people do Hint t statistics and p values are different ways of looking at the same issue A t sta tisti
248. ing and make adjustments if needed hit You can click in each column to change the series name or enter a description for the series In our example EViews has correctly analyzed the file so we can just hit and EViews generates our workfile EViews intuition is pretty good when it comes to reading Excel files so fre quently the first step is also the step Sometimes though we have to lend a hand The file Treasury_Interest_Ra tes xls on the EViews web site pro vides a few examples If you drag and drop Treasury_Interest_Ra tes xls onto EViews the Spreadsheet read EViews Illustrated book Page 41 Monday February 25 2013 10 06 AM 42 Chapter 2 EViews Meet Data dialog opens to let us choose which sheet to read from the file It so happens we want the second sheet Monthly which we can choose in the Predefined range dropdown The controls in the upper left hand corner of the second Spreadsheet read dialog provide a number of options for customizing how EViews interprets the Excel spreadsheet Which option you need depends on how your file is structured Here s one example An excerpt from our Excel file is shown to the right This particular file has a description of each variable in the first row and the name of the variable in the second row Since the default assumption is that only the variable name is present this won t do We need to provide more information The Hea
249. ings of text Once in a while data ain t In other words sometimes you just don t know the value for a particular data point so you mark it NA How do you tell EViews that a particular observation is not available If you re entering data by typing or copy and paste you don t have to tell EViews EViews initializes data to NA If you don t know a particular value leave it out and it will remain marked NA The harder issue comes when you re reading data in from existing computer files There are two separate issues you may have to deal with How do you identify NA values to EViews What if multiple values should be coded NA Reading NAs from a file There are a couple of situations in which EViews identifies NAs for you automatically First if EViews comes across any nonnumeric text when it s looking for a number EViews con verts the text to NA For example the data string 1 NA 3 will be read as the number 1 an NA and the number 2 The string 1 two 3 will be read the same way there s nothing magic about the letters NA when they appear in an external file Second EViews will usu ally pick up correctly any missing data codes from binary files created by other statistical programs If you have a text file or an Excel file which has been coded with a numerical value for NA 999 and 0 are common examples you can tell EViews to translate these into NA Statistical hint
250. ir visual appearance Area Graph An area graph is a line graph with the area underneath the line filled in The same information is dis played in line and area graphs but area graphs give a sense that higher values are bigger Interest rates are probably better depicted as line graphs In contrast an area graph of the federal debt held by the public emphasizes that the U S national debt is one whole heck of a lot more than it used to be EViews Illustrated book Page 132 Monday February 25 2013 10 06 AM Picture One Series 133 Bar Graph A bar graph represents the height of each point with a vertical bar This is a great format for display ing a small number of observa tions and a crummy format for displaying large numbers of obser vations The figure to the right shows federal debt for the first observation in each decade Note that EViews has drawn vertical lines to indicate breaks in the sample Bar labels can be added with the click of a radio button in the Fill Areas tab of the Graph Options dialog An example is shown to the right Note that we have also used the options page to add a neat fade effect to the bars EViews Illustrated book Page 133 Monday February 25 2013 10 06 AM 134 Chapter 5 Picture This Spike Graph A spike graph is just like a bar graph only with really skinny bars It s especially useful when you have too many categories to display neatly with a bar graph Her
251. is Here s a simple pie chart con sumption sectors wf1 showing the relative sizes of the con sumption of durables nondura bles and services in the United States in 1959 and 1999 We ve turned on Label Pies in the Bar Area Pie section of the Graph Elements group in the Graph Options dialog It s easy to see here the dramatic increase in the size of the service sector Most EViews graphs render pretty nicely in monochrome even if you ve created the graphs in color Pie graphs don t make out quite so nicely so you ll want to do a little more custom ization for black and white images The aesthetic problem is that pie graphs have large filled in areas right next to each other Colors distinguish in monochrome adjacent filled areas don t look so good If you have access to both the black and white printed version of this book and the electronic color version compare the appearance of the pie chart above Color looks a lot better The solution is further complicated because there are two ways to get a monochrome graph out of EViews You can tell EViews to render it without color by unchecking Use Color in the appropriate Save or Copy dialogs or by sending it directly to a monochrome printer When EViews knows that it s rendering in black and white it does a decent job of picking grey scales Alternatively you can copy the image in color into another program and then let the other program render it the best i
252. is NA in which case it returns Y For example to recode NAs in X to 999 use X nan X 999 Auto Series and Two Examples Pretty much any place in EViews that calls for the name of a series you can enter an expres sion instead EViews calls these expressions auto series Showing an expression For example to check on our use of recode on page 91 you can enter an expression directly in a show command thusly show one_2_3 two_3_1 recode one_2_3 gt two_3_1 one_2_3 two_3_1 A foolish consistency is the hobgoblin of little minds hint Logically the result of the com parison x na should always return NA in line with the rule that any operation involving an NA results in an NA Logical perhaps but use less EViews favors common sense so this operation gives the desired result Q Can I define my own function A No EViews Illustrated book Page 92 Monday February 25 2013 10 06 AM Your Basic Elementary Algebra 93 Auto series in a regression Here s an example which illustrates the econometric theorem that a regression includ ing a constant is equivalent to the same regression in deviations from means exclud ing the constant We can use the random number generators to fabricate some data rndseed 54321 series x rnd series y 2 3 x nrnd Then we can run the usual regression with ls y c x The results are as expected Both the intercept and slope are close to the values that we
253. is to import data that someone else has already entered into a computer file But let s assume that we re going to type the data in from scratch Table 1 Academic Salary Data Excerpt dis plays three variables You have a choice of entering one variable at a time or entering several in a table format We ll illustrate both methods Suppose first we re going to enter one series at a time starting with NONACADSAL We want to create a new series and then fill in the appropriate values The trick is to open a window with an empty series and then fill it up There are a bunch of ways to get the desired window to pop open These two methods pop open an Untitled series b EViews Illustrated book Page 27 Monday February 25 2013 10 06 AM 28 Chapter 2 EViews Meet Data Type the command series in the command pane Use the menu commands Object New Series These two methods create a series named NONACADSAL and place it in the workfile Type the command series nonacadsal in the command window Use the menu commands Object New Series and then enter NONACADSAL in the Name for object field The latter two methods place in the workfile Double click to open a series window In contrast the former two methods open a window automatically but don t name it These methods open an untitled series win dow To name the untitled series click on the button and enter NONACADSAL EViews doesn t
254. is a separate graphic sub object You can set options for the graphs individually or all together You can also grab the graphs with the mouse and drag them around to re arrange their locations in the window EViews Illustrated book Page 138 Monday February 25 2013 10 06 AM Group Graphics 139 Stack Em High Several graph types let you stack multiple series which is sort of like adding the data val ues in the series vertically The first series is plotted in the usual way The second is plotted as the sum of the first and the second series The third as the sum of the first three series And so forth Here s an ordinary bar graph from the workfile con sumption sectors wf1 showing the various pieces of U S con sumption in 1999 Hint If you prefer EViews can auto arrange the individual graphs into neat rows and columns Right click on the graph window and choose Position and align graphs EViews Illustrated book Page 139 Monday February 25 2013 10 06 AM 140 Chapter 5 Picture This Using the Multiple series field to tell EViews to Stack in single graph gives a better sense of how much total consumption was as well as packing in the information while using less space Several graph types pro vide this kind of stacking option Other than deciding on window arrangements Line Area Bar and Spike graphs are the same for a group as for a series except that you get one line set of bars etc
255. is interpreted as the number of days since Monday January 1 0001 CE Date series are manipulated in three ways you can control their display in spreadsheet views you can convert back and forth between date numbers and their text representation and you can perform date arithmetic Date Displays Open a series T 0 0 5 1 1 5 2 and you get the standard spreadsheet view Use the button to bring up the Properties dialog Choose Day Time to change the display to treat the numbers as dates showing both day and time EViews Illustrated book Page 106 Monday February 25 2013 10 06 AM Can We Have A Date 107 The display changes as shown to the right Notice that fractional parts of numbers correspond to a fraction of a day Thus 1 5 is 12 noon on the second day of the Common Era Two Day and Time formats are also shown by way of illustration The Date format dropdown menu provides a variety of date display formats Date to Text and Back Again January 1 1999 is more easily understood by humans than is its date number representation 729754 On the other hand comput ers prefer to work with numbers A variety of functions translate between numbers inter preted as dates and their text representation The function datestr x fmt translates the number in X into a text string A second optional argument specifies a format for the date As examples datestr 731946 pro duces 1 1 2005 datestr 731946
256. is it That s the message of this little subsec tion Many users live fulfilled and truly happy lives without ever messing with the Options menu This is okay EViews designers have chosen very nice defaults for the options Feel free to leave them as they are Option al Recommendations Here s how you should reset your options Trust me We ll talk about why later These options can be found under the General Options menu item Recapitulation note Parts of the discussion here repeat advice given in earlier chap ters EViews Illustrated book Page 403 Monday February 25 2013 10 06 AM 404 Chapter 18 Optional Ending Window Behavior In the Environ ment Window behav ior dialog under Warn on close uncheck Series Matrices Coeffi cients Groups Tables Graphs and Equation Sys VAR Pool Model Under Allow only one untitled uncheck every thing to cut down on unnecessary alert boxes Alpha Truncation In the Series and Alphas Alpha trunca tion dialog enter a large number in the Maximum number of characters per obser vation field Try 256 or even 1 000 EViews Illustrated book Page 404 Monday February 25 2013 10 06 AM Option al Recommendations 405 Workfile Storage Defaults In the Data stor age Workfile Save dia log check Use compression One exception don t do this if you need to share workfiles with someone using an older before 5 0 version of EViews Uncheck Prompt on eac
257. is that there are lots of population and output series one for each country We use pools to study behavior common to all the countries The second thing you ll notice is that series names have two parts a series component identifying the series and a cross section component identifying the cross section element the country in this example So POPCAN is population for Canada and POPFRA is population for France YCAN is Canadian output and YFRA is French output There s just one rule you have to remember about series set up in a pool Pooled series aren t any different from any other series they re simply ordinary series conveniently named with common components In other words pool series have neither any special features nor any special restrictions The only thing going on is that their names are set up conveniently to identify the country or other cross sectional element with which they re associated For example the com mand EViews Illustrated book Page 291 Monday February 25 2013 10 06 AM 292 Chapter 12 Everyone Into the Pool ls yfra c d log popfra gives us the regression of output on population growth for France The reported effect of population growth is statistically significant and rather large Given historical magnitudes in French rates of pop ulation growth the effect accounts for a decrease in output of about 10 percent relative to US output Into the Pool Pooled series aren
258. ita Y is measured relative to U S GDP Stacked Means Removed Specifying Stacked means removed in Descriptive Statistics produces some pretty funny looking output but it turns out that this method is just what we want for answering certain questions EViews subtracts the means for each country before generating the descriptive statistics As a consequence the means are always zero which looks pretty funny The raison d tre for Stacked means removed is to see statistics other than the means and medians The medians aren t zero but they re pretty close According to the Stacked data statistics the standard deviation of annual U S population growth was 3 10ths of one percent while the standard deviation for the pooled countries was 6 10ths of one percent This looks like population growth was much more variable for the countries in the pooled sample Whether this is the correct conclusion depends on a subtle point Some of the countries have rela tively high population growth and some have lower growth The standard deviation for Stacked data includes the effect of variability across countries and across time while the standard deviation reported for the United States is looking only at variability across time The Stacked means removed report takes out cross country variability reporting the time EViews Illustrated book Page 299 Monday February 25 2013 10 06 AM 300 Chapter 12 Everyone Into the Pool seri
259. ked View Re arranging the spreadsheet into the usual order that is by date is called the unstacked view Clicking the button flicks back and forth between stacked and unstacked views Pooled Statistics The Descriptive Statistics view offers a number of ways to slice and dice the data in the pool We ve put two pooled series with the marks and one non pooled series in the dialog so you can see what happens as we try out each option First look at the Sample radio buttons on the right The presence of missing data NAs means that the samples available for one series may differ from the sample available for another You can see above for example that Canada France and Great Britain have data starting in 1950 but that German data begins later Common sample instructs EViews to use only those observations available for all countries for a particular series while Balanced sample requires observations for all countries for all series entered in the dialog Individual sample means to use all the observations available EViews Illustrated book Page 298 Monday February 25 2013 10 06 AM Playing in the Pool Data 299 Stacked Data Statistics The default Data Organization is Stacked data which stacks the series for all countries together for the purpose of producing descriptive statistics For example we see that GDP per capita in our six pooled countries averaged just over 70 percent of U S GDP per cap
260. know whether the difference is meaningful or whether it s a random sta tistical fluke Formally we want to test the hypothesis that the mean University of Washington score equals 152 and ask whether there is sufficient evidence to reject this hypothesis We will perform this test after setting the sample to include observations where the GPA is within normal bounds and the LSAT score exceeds 100 Choose Descriptive Sta tistics amp Tests and Simple Hypothesis Tests and then enter the hypothesized mean in the Series Distribution Tests dialog Hint If you re only testing a mean you don t need to fill out any other fields The Variance and Median fields are for hypothesis tests on the variance or median In par ticular don t enter a previously estimated standard deviation in the Enter s d if known field That s only for the unusual case in which a standard deviation is known rather than having been estimated EViews Illustrated book Page 211 Monday February 25 2013 10 06 AM 212 Chapter 7 Look At Your Data EViews conducts a standard t test for this hypothesis providing both the t statistic and its associated p value In this case the p value tells us that if the true average LSAT in the appli cant pool was 152 the probability of observing the mean LSAT found in our data 158 is zero to four decimal places Clearly applicants to the Uni versity of Washington s very good law school are better than the a
261. l Ending Font Options The Fonts dialog con trols default fonts in the workfile display in spreadsheets and in tables Font selection is an issue about what looks good to you so turn on whatever turns you on Frequency Conversion The Series and Alphas Frequency conversion dialog lets you reset the default frequency conversions Usually there s a way to control the conversion method used for indi vidual conversions Sometimes copy and paste for example there isn t See Multiple Frequen cies Multiple Pages in Chapter 9 Page After Page After Page for an extended discussion of frequency conversion EViews Illustrated book Page 408 Monday February 25 2013 10 06 AM Alpha Truncation 409 Alpha Truncation EViews sometimes truncates text in alpha series You probably don t want this to hap pen Unless you re stor ing large amounts of data in alpha series increase the maximum number of characters so that nothing ever gets truncated With modern computers large amounts of data means on the order of or observa tions Spreadsheet Defaults Aside from the com ments in Option al Rec ommendations above there s really nothing you need to change in the Spreadsheets defaults section Look ing at the Layout page you may want to check one or more of the Edit mode on checkboxes If you do then the corre sponding spreadsheets open with editing per mitted Leaving
262. last ten or so changes Typing a Table at a Time Now let s turn to entering data in the form of a table As an example we ll enter the name of the academic discipline and the academic salary together We begin with the same choice do we name the series before or after we open the window If you like to name first do the following Type the following commands in the command window series salary alpha discipline Select DISCIPLINE and then SALARY in the workfile window hold down the Ctrl key to select both series double click and choose Open Group to open a group window Hint It s worth the trouble to add as much documentation as possible in the series label Later you ll be glad you did Hint Series in a group window are displayed from left to right in the same order as you click on them in the workfile window EViews Illustrated book Page 30 Monday February 25 2013 10 06 AM Time to Type 31 Note that the series DISCIPLINE is displayed with the icon to signal a series holding alphabetic data as contrasted with the icon for ordinary numeric series If you like to open a window before creating a series do the following Use the menu Quick Empty Group Edit Series When the window opens scroll up one line Then type DISCIPLINE in the cell next to the cell marked A dialog pops up so that you can tell what sort of series this is going to be Since DISCIPLINE is text rather than numbers
263. le The Basic EViews Document 3 Viewing an individual series 4 Looking at different samples 6 Generating a new series 8 Looking at a pair of series together 11 Estimating your first regression in EViews 13 Saving your work 18 Forecasting 20 What s Ahead 21 CHAPTER 2 EVIEWS MEET DATA 23 The Structure of Data and the Structure of a Workfile 24 Creating a New Workfile
264. lements of C equal the just estimated coefficients In fact every time EViews estimates an equation it stores the results in the vector C The key to nonlinear estimation is Feed the ls command a formula in place of a list of series If you enter the command ls volume c 1 c 2 trend Hint It may look curious that C 1 C 2 etc seem to be labeled R1 R2 and so on R1 and R2 are generic labels for row 1 and row 2 in a coefficient vector The same setup is used generally for displaying vectors and matrices volume a b trend a b EViews Illustrated book Page 341 Monday February 25 2013 10 06 AM 342 Chapter 14 A Taste of Advanced Estimation you get precisely the same results as above The only difference is that the formula in the command is reported in the top panel and that C 1 and C 2 appear in place of series names Naming Your Coefficients Least squares with a series list always stores the estimated coefficients in C but you re free to create other coefficient vectors and use those coefficients when you specify a formula To create a coefficient vector give the command coef n newname replacing n with the desired length of the coefficient vector and newname with the vector s name Here s an example coef 1 alpha coef 2 beta ls volume alpha 1 beta 1 trend Hint Unlike series where a number in parentheses indicates a lead or lag the number following a coefficien
265. ls 335 338 wfcreate command 25 26 White heteroskedasticity test 339 windows locking unlocking 29 within plot 158 wls weighted least squares 335 338 workfile samples 95 workfiles adding data 27 29 backing up 243 399 characteristics 3 4 239 creating 25 26 dated 34 37 loading 3 moving within 29 restructuring 33 saving 18 19 storage options 405 410 411 structure 25 27 X XY line graphs 149 152 Y year function 91 yield curve 130 EViews Illustrated book Page 424 Monday February 25 2013 10 06 AM
266. ltiply by 365 Or we could use the more precise formula that takes compounding into account But some of our returns accumulated over a weekend and arguably represent three days earnings Taking into account that the number of days between observations varies we can annualize the return with series return 365 datediff tradedatenum tradedate num 1 dd pct_change Let s take this apart The expression datediff tradedatenum tradedatenum 1 dd returns the number of days between observations Usually there s one day between obser 1 pct_change 365 1 EViews Illustrated book Page 108 Monday February 25 2013 10 06 AM What Are Your Values 109 vations giving us 365 1 Over an ordinary weekend the datediff function returns 3 so we annualize by multiplying by 365 3 Note two things about annualized returns First typical daily returns imply very large annual rates of change In fact the annual rates are implausible Second we ve captured not only weekends but also the January 17th closing in honor of Dr King What Are Your Values The workfile CPSMAR2004Extract wf1 is a cross section of indi viduals from the Current Population Survey The series FE codes a person s gender As you will remember from early biology les sons humans have two genders 0 and 1 Some of us prefer to think of humans as male and female rather than 0 and 1 EViews uses value maps to translat
267. lways analyzes multiple series together as a group To get started EViews Illustrated book Page 117 Monday February 25 2013 10 06 AM 118 Chapter 5 Picture This create a group by selecting the three month one year and 20 year interest rate series TM3 TY01 and TY20 with the mouse opening them as a group and then naming the group RATES_TO_GRAPH As a reminder you select multiple series by holding down the Ctrl key Equivalently type the command group rates_to_graph tm3 ty01 ty20 and then double click to open the group The menu View Graph brings up the Graph Options dialog This mas ter dialog can be used to create a wide variety of graph types and also provides entry for tuning a graph s appear ance after it s cre ated For now hit to produce a simple line graph This graph looks pretty good You can print it out or copy it into a word pro cessor and be on your merry way EViews Illustrated book Page 118 Monday February 25 2013 10 06 AM A Simple Soup To Nuts Graphing Example 119 Get A Closer Look Notice the slider bar at the bottom of the graph If we want to look more closely at a specific part of the data we can resize it and move it to blow up a part of the graph To see the action during the peak we drag and resize the slider bar to show data between 1978 and 1985 You should be aware that when you close the window the visible range will be reset If you d like
268. ly dumb ones series two_seconds two_seconds two_seconds a really dumb command Hint EViews regards text following an apostrophe as a comment that isn t pro cessed a really dumb command is a note for humans that EViews ignores Hint There s no Undo command Once you ve replaced values in a series they re gone Moral Save or SaveAs frequently so that if necessary you can load back a pre mistake version of the workfile EViews Illustrated book Page 84 Monday February 25 2013 10 06 AM Your Basic Elementary Algebra 85 The series command performs two logically separate operations It declares a new series object TWO_SECONDS Then it fills in the values of the object by computing ONE_SECOND ONE_SECOND We could have used two commands instead of one series two_seconds creates a series in the workfile named TWO_SECONDS initialized with NAs We could then type two_seconds one_second one_second Once a series has been created or declared in computer speak the command name series is no longer required at the front of a data transformation line although it doesn t do any harm Some Typing Issues The command pane provides a scrol lable record of commands you ve typed You can scroll back to see what you ve done You can also edit any line including using copy and paste to help on the editing Hit Enter and EViews will copy the line containing the cursor to the botto
269. m of the command pane and then exe cute the command You may wish to use CTRL UP to recall a list of previous commands in the order they were entered The last command in the list will be entered in the command window Holding down the CTRL key and pressing UP repeatedly will display the prior commands Repeat until the desired com mand is recalled for editing and execution If you ve been busy entering a lot of commands you may press CTRL J to examine a his tory of the last 30 commands Use the UP and DOWN arrows to select the desired command and press ENTER or double click on the desired command to add it to the command win dow To close the history window without selecting a command click elsewhere in the command window or press the Escape ESC key Hint EViews doesn t care about the capitalization of commands or series names EViews Illustrated book Page 85 Monday February 25 2013 10 06 AM 86 Chapter 4 Data The Transformational Experience The size of the command pane is adjustable Use the mouse to grab the separator at the bot tom of the command pane and move it up or down as you please You may also drag the command window to anywhere inside the EViews frame Press F4 to toggle docking or click on the command window depress the right mouse button and select Toggle Command Docking You can print the command pane by clicking anywhere in the pane and then choosing File Print Similarly you can save the comman
270. marily as a safety check A quick glance at the mean of the dependent variable guards against forgetting that you changed the units of measurement or that the sample used is somehow different from what you were expecting Adjusted R squared makes an adjustment to the plain old to take account of the num ber of right hand side variables in the regression measures what fraction of the varia tion in the left hand side variable is explained by the regression When you add another right hand side variable to a regression always rises This is a numerical property of least squares The adjusted sometimes written subtracts a small penalty for each additional variable added F statistic and Prob F statistic come as a pair and are used to test the hypothesis that none of the explanatory variables actually explain anything Put more formally the F sta R2 R2 R2 R2 R 2 EViews Illustrated book Page 72 Monday February 25 2013 10 06 AM A Multiple Regression Is Simple Too 73 tistic computes the standard F test of the joint hypothesis that all the coefficients except the intercept equal zero Prob F statistic displays the p value corresponding to the reported F statistic In this example there is essentially no chance at all that the coefficients of the right hand side variables all equal zero Our final summary statistic is the Durbin Watson the classic test statistic for serial corre
271. mathematical functions such as sin x cos x mean x median x max x var x All the functions take a series as an argument and produce a series as a result but note that for some functions such as mean x the output series holds the same value for every observation Random Numbers EViews includes a wide variety of random number generators See Statistical Functions Three functions for generating random numbers that are worthy of special attention are rnd uniform 0 1 random numbers nrnd standard normal random numbers and rndseed set a seed for random number generation Officially these functions are called special expressions rather than functions Statistical programs generate pseudo random rather than truly random numbers The sequence of generated numbers look random but if you start the sequence from a particular value the numbers that follow will always be the same Rndseed is used to pick a starting point for the sequence Give rndseed an arbitrary integer argument Every time you use the same arbitrary argument you ll get the same sequence of pseudo random numbers This lets you repeat an analysis involving random numbers and get the same results each time Hint log means natural log To quote Davidson and MacKinnon s Econometric Theory and Methods In this book all logarithms are natural logarithms Some authors use ln to denote natural logarithms and log to denote
272. me task with the recode command as in x recode x 9 or x 99 or x 999 NA x If the logical condition in the first argument of recode is true X is missing in this exam ple the value of recode is the second argument NA otherwise it s the third argument X Quick Review The easiest way to get data into EViews is to read it in from an existing data file EViews does a great job of interpreting data from spreadsheet and text files as well as reading files created by other statistical programs Hint It might be wiser to make a new series say XRECODE rather than change X itself This leaves open the option to treat the different missing codes differently at a later date If you change X there s no way later to recover the distinct 9 99 and 999 codes EViews Illustrated book Page 57 Monday February 25 2013 10 06 AM 58 Chapter 2 EViews Meet Data Whether reading from a file or typing your data directly into an EViews spreadsheet think of the data as being arranged in a rectangle observations are rows and series are columns Appendix Having A Good Time With Your Date EViews uses dates in quite a few places Among the most important are Labeling graphs and other output Specifying samples In data series Most of the time you can specify a date in any reasonable looking way The following com mands all set up the same monthly workfile wfcreate m 1941m12 1942m1 wfcreate m 41 12
273. mention that discrepancy thing In the equation view EViews displays only enough of each equation so that you can remember which equation s which To see the full equation switch to the text view or look at the equation s Prop erties Hint The estimated standard error is used when you ask EViews to execute a stochas tic simulation a feature we won t explore referring you instead to the User s Guide EViews Illustrated book Page 373 Monday February 25 2013 10 06 AM 374 Chapter 15 Super Models Numerical accuracy Computers aren t nearly so bright as your average junior high school student so they use numerical methods which come up with approximate solutions If you d like a more accurate answer you need to tell EViews to be more fussy Click and choose the Solver tab Change Convergence to 1e 09 to get that one extra digit of accuracy Your Second Homework Odds are that your second homework assignment in your first introductory macroeconomics class asked what would happen to GDP if G were to rise In other words how do the results of this new scenario differ from the baseline results Vanity hint In the problem at hand all we re doing is making the answer look pretty In more complicated problems a smaller convergence limit has the advantage that it helps assure that the computer reaches the right answer The disadvantages are that the solution takes longer and that sometimes if you ask f
274. menus 9 11 86 method 70 mixed frequency data 244 251 mle maximum likelihood estimation 355 357 models accuracy of solution 374 ARIMA 329 333 ARMA 329 333 baseline data 375 376 378 creating 366 features 380 first order serial correlation 315 forecasting dynamic systems of equations 378 379 GARCH 353 logit 350 351 scenarios 375 376 solving 366 368 uses 365 variables 372 377 viewing solution 370 371 monochrome graphs 154 155 Monte Carlo 386 389 month function 91 moving average MA errors 322 323 324 325 See also autoregressive moving average ARMA errors multiple graphs 127 137 multiple regression 73 74 N NA values 56 57 69 70 87 92 98 103 named auto series 94 95 103 named objects 407 naming objects 18 28 29 NAN function 92 near outliers 209 nonlinear least squares estimation 341 345 nonlinear models 380 normal distributions 114 199 now function 108 number loops 385 numbers data type 100 102 display of 101 102 displaying meaningful 65 separators 406 n way tabulation 219 221 O object views 398 399 objects commands 398 399 creating system 357 creating VAR 361 defined 5 freezing 393 394 EViews Illustrated book Page 420 Monday February 25 2013 10 06 AM Q 421 generally 398 399 graph 120 121 named 407 naming 18 28 29 paste pictures into inability to 126 updating 400 views 398 399 obs series 32 observation number
275. mmand smpl all in the command pane and end with the Enter key Hint Almost everything in EViews can be done either by typing commands or by choosing a menu item The choice is a matter of personal preference EViews Illustrated book Page 9 Monday February 25 2013 10 06 AM 10 Chapter 1 A Quick Walk Through You can see that the sample in the workfile window has changed back to 1888Q1 through 2004Q1 Now that you ve set the sample to include all the data let s generate LOGVOL again this time from the command line Type series logvol log volume in the command pane and hit Enter This is the last time I ll nag you about hitting the Enter key I promise Historical hint Ever wonder why so many computer commands are limited to four let ters Back in the early days of computing several widely used computers stored char acters four bytes to the word It was convenient to manipulate data a word at a time Hence the four letter limit and commands spelled like smpl EViews Illustrated book Page 10 Monday February 25 2013 10 06 AM Looking at a pair of series together 11 Again double click on LOGVOL to check that we now have all our data Then use the View menu to choose View Graph and select Line graph The line graph for LOGVOL the logarithm of our original VOLUME variable appears What we see is not quite a straight line but it s a lot closer to a straight line and a lot
276. more But you knew that Hint As a general rule different groups of data summarized in a single plot need to be commensurable meaning they should all have the same units of measurement Our three interest series are all measured in percent per annum In contrast even though GDP and unemployment are both indicators of economic activity it makes no sense to compare a mean measured in billions of dollars per year with a mean measured in per centage points Hint Details only works for some graph types For types where it doesn t the Details option won t appear EViews Illustrated book Page 157 Monday February 25 2013 10 06 AM 158 Chapter 5 Picture This Factoring out the categories EViews allows multiple categorical variables each with multiple categories This can mean lots and lots of individual plots EViews uses the Factors series defining categories field to sort out which variables place plots within a graph and which place plots across graphs For the graph above we put UNION in the Within graph box This told EViews to place the bars for all the categories of UNION nonunion and union within the same graph In contrast if we d entered UNION in the Across graphs box EViews would have spread the bars across plots so that each UNION category appears separately as below Splitting up wages by gender as well as union membership we have four plots with numer ous arrangement possibilities I
277. n elements set up a panel Hint The similarities between pools and panels are greater than the differences and in any event it s not hard to move back and forth between the two forms of organiza tion EViews Illustrated book Page 272 Monday February 25 2013 10 06 AM Quick P review 273 Now that you ve had a quick taste proceed to Chapter 11 Panel What s My Line and Chapter 12 Everyone Into the Pool to get the full flavor EViews Illustrated book Page 273 Monday February 25 2013 10 06 AM 274 Chapter 10 Prelude to Panel and Pool EViews Illustrated book Page 274 Monday February 25 2013 10 06 AM Chapter 11 Panel What s My Line Time series data typically provide one observation in each time period annual observations of GDP for the United States would be a classic example In the same way cross section data provide one observation for each place or person We might for example have data on 2004 GDP for the United States Canada Grand Fenwick etc Panel data combines two dimensions such as both time and place for example 30 years of GDP data on the United States and on Canada and on Grand Fenwick Broadly speaking we want to talk about three things in this chapter First we ll talk about why panel data are so nifty Next comes a discussion of how to organize panel data in EViews Finally we ll look at a few of EViews special statistical procedures for panel data What s So
278. n should change the standard errors However the estimated coefficients shouldn t change by very much Technically both the original and corrected results are unbiased In our example the coefficient on D LOG CLOSE 1 went from positive and significant to negative and insignificant This is an informal signal that the dynamics in this equation weren t well specified in the original estimate Higher Order Corrections Correcting for higher order autoregressive errors and for moving errors is just about as easy as correcting for an AR 1 once you understand one very clever notational oddity EViews requires that if you want to estimate a higher order process you need to include all the lower order terms in the equation as well To estimate an AR 2 include AR 1 and AR 2 To estimate an AR 3 include AR 1 AR 2 and AR 3 If you want an MA 1 include MA 1 in the regression specification And as you might expect you ll need MA 1 and MA 2 to estimate a second order moving average error Why not just type AR 2 for an AR 2 Remember that a second order autoregression has two coefficients and If you type AR 1 AR 2 both coefficients get estimated Omitting AR 1 forces the estimate of to zero which is something you might want to do on rare occasion probably when modeling a seasonal component Autoregressive and moving average errors can be combined For example to estimate both an AR 2 and an
279. n that we re necessarily faking the data a little There isn t any monthly GDP data All we re doing is taking a reasonable guess In contrast high to low frequency conversion doesn t involve making up data at all If we copy IP from the monthly Indpro page and paste it into the quarterly Gdpc96 page we see the following EViews Illustrated book Page 249 Monday February 25 2013 10 06 AM 250 Chapter 9 Page After Page After Page EViews has applied the default high to low frequency conver sion procedure which is to average the monthly observations within each quarter The meanings of the high to low frequency conversion options are pretty straightforward Sum observa tions adds up the monthly observations within a quarter If we had sales for January February and March we would use a Sum observations conversion to get total first quarter sales First Last Max and Min Observation all pick out one month within the quarter and use the value for the quarterly value Default Frequency Conversions Every series has built in default frequency conversions one for high to low and one for low to high These defaults are used when you copy and paste from one page or workfile to another To change the default choices open the series and click the button Choose the Freq Convert tab to access available choices The overall EViews default is changed using the menu Options General Options Series and Alphas Frequency Conversion
280. nce A Rolling Example Here s a very practical problem that s easily handled with a short program The workfile currency rolling wf1 has monthly data on currency growth in the variable G The com mands smpl all ls g c g 1 fit gfoverall estimate a regression and create a forecast Of course this example uses current data to esti mate an equation which is then used to forecast for the previous century We might instead perform a rolling regression in which we estimate the regression over the previous year use that regression for a one year forecast and then do the same for the following year The program rolling_currency prg shown to the right does what we want Note the use of a for loop to control the sample a common idiom in EViews Note also that we re used the same equation object for each regression to avoid opening multiple windows Hint Prefacing a line in a program with the word do also suppresses windows being opened do doesn t affect object creation EViews Illustrated book Page 387 Monday February 25 2013 10 06 AM 388 Chapter 16 Get With the Program For this particular set of data the rolling forecasts are much closer to the actual data than was the overall forecast Quick Review An EViews program is essentially a list of commands available to execute as needed You can make the commands apply to different objects by providing arguments when you run the program You
281. nday February 25 2013 10 06 AM 102 Chapter 4 Data The Transformational Experience mals This can be especially useful for financial data in which prices were required to be rounded For example 10 1 8 rather than 10 125 EViews saves any display changes you make in the series window Properties dialog Dis play tab and uses them whenever you display the series Remember that the Properties dia log changes the way the number is displayed not the underlying value of the number Letters EViews second major data type is text data stored in an alpha series displayed with the icon You create an alpha series by typing the command alpha aname or using the Object New Object Series Alpha menu Double clicking an alpha series opens a spreadsheet view which you can edit just as you would a numeric series Hint To change the default display use the menu Options Spreadsheet Defaults See Chapter 18 Optional Ending EViews Illustrated book Page 102 Monday February 25 2013 10 06 AM Numbers and Letters 103 Alpha series have two quirks that matter once in a great while The first quirk is that the not available code for alpha series is the empty string rather than NA Visually it s difficult to tell the difference between an empty string and a string holding nothing but one or more spaces They both have a blank look about them The second quirk is that all alpha series in EViews
282. ne or transform it into pooled form using Proc Reshape Current Page Unstack in New Page and filling out the Workfile Unstack dialog as shown Click and you have a new page set up in pooled form In fact to help out EViews has even set up a pool object for you EViews Illustrated book Page 311 Monday February 25 2013 10 06 AM 312 Chapter 12 Everyone Into the Pool Exporting Stacked Data Typically unstacked data are easier to operate on but sometimes stacked data are easier for humans to read The inverse of Proc Import Pool data ASCII XLS WK is Proc Export Pool data ASCII XLS WK Choose By Date or By Cross section and away you go Exporting Stacked Data A Little Indirection Here Too Not surprisingly there s an indirect method for exporting too To stack data in a new page in preparation for using any of the usual export tools choose Proc Reshape Current Page Stack in New Page In the Workfile Stack dia log enter the name of the pool object Click for a nicely stacked page Quick Review Pools The pool feature lets you analyze multiple series observed for the same variable such as GDP series for a number of countries You can pool the data in a regression with common Hint EViews includes the in the series name in the output file You might choose to manually delete the in the exported file to improve the appearance of the out put EViews Illustrated book Pa
283. ne that population growth is picking up the effect of omitted variables that we can t measure To the extent that the omitted variables are constant for each country fixed effects estimation will control for the omissions Setting the sample back to everything except the United States click the but ton and then choose the tab Set Effects specification to Cross sec tion Fixed This instructs EViews to include a separate intercept for each coun try Econometric digression The regression output includes a hint that something funky is going on The Durbin Watson statistic see Chapter 13 Serial Correlation Friend or Foe indicates very very high serial correlation This suggests that if the error for a country in one year is positive then it s positive in all years and if it s negative once then it s always negative High serial correlation in this context provides a hint that we ve left out country specific information ai EViews Illustrated book Page 281 Monday February 25 2013 10 06 AM 282 Chapter 11 Panel What s My Line In our new regression results the effect of population growth is reduced to about one 1 100th of the previous estimate This confirms our suspicion that the previous estimate had omitted variables and apparently ones that mattered a lot We can take a look at the estimated values of the fixed effects for each country by looking at the Fixed Ran dom Effects Cross secti
284. new page Choosing Specify by Identifier Series brings up the Workfile Page Create by ID dialog Enter GMSTCEN in the Cross ID series field It s optional but we ve also entered a name for the new page in the Page field at the lower right EViews Illustrated book Page 259 Monday February 25 2013 10 06 AM 260 Chapter 9 Page After Page After Page The new page opens containing just the series GMSTCEN Okay the new page also contains FM11X but that s only because FM11X is a value map holding the names of each state If we double click on GMSTCEN we see that GMSTCEN has also supplied the identifier series for this page which appears in the left most shaded column of the spreadsheet Now that we have a page identified by state we need to fill it up with state by state data One easy method is copy and paste Go back to the individual data page Cps Ctrl click on LNWAGE ED AGE and UNION to select the relevant series Copy and then click back on the ByState tab Choose Paste from the context menu Because Paste and Paste Special are the same here this brings up the Paste Special dialog We ll discuss the Match merge options further in a bit but EViews has done it s usual good job of guessing what we want done For now just note that the field Con traction method is set to Mean and hit the button EViews Illustrated book Page 260 Monday February 25 2013 10 06 AM Matching When The Identifiers Are
285. ng hypothesis tests Measures of how well the equation fits the data How much is explained by the esti mated values of and and how much remains unexplained Diagnostics to check up on whether assumptions underlying the regression model seem satisfied by the data We re re using the data from Chapter 1 A Quick Walk Through to illustrate the features of EViews regression procedure If you want to follow along on the computer use the workfile NYSEVOLUME as shown a b volumet log 2 629649 0 017278t a b ut e volumet log a bt ut b 0 017 a b EViews Illustrated book Page 62 Monday February 25 2013 10 06 AM A First Regression 63 EViews allows you to run a regression either by creating an equation object or by typing commands in the command pane We ll start with the former approach Choose the menu command Object New Object Pick Equation in the New Object dialog The empty equation window pops open with space to fill in the variables you want in the regression Regression equations are easily specified in EViews by a list in which the first variable is the dependent variable the vari able the regression is to explain followed by a list of explana tory or independent vari ables Because EViews allows an expression pretty much any where a variable is allowed we can use either variable names or expressions in our re
286. nge all sample Herodotus first 2000 sample Heinlein 2001 last Using the command line we create our forecasts with smpl herodotus ls g trend g 1 expand month smpl heinlein fit ghein_stat forecast ghein_dyn plot g ghein_stat ghein_dyn After a little touch up our graph looks like this The static and dynamic forecasts look similar and track actual currency growth well So using this model to fore cast the real future seems prom ising int not intended for Americans Unless you were born in earshot of Bow bells in which case it s oleRange erodotus and einlein not that h EViews will h under stand EViews Illustrated book Page 230 Monday February 25 2013 10 06 AM Facing the Unknown 231 Setting the Sample in the Forecast Dialog The forecast dialog can be used if you prefer it to typing commands Enter the forecast sample HEIN LEIN in the Forecast sample field By default Insert actuals for out of sample observations is checked Under the default EViews inserts observed G into GHEIN_DYN for data points that aren t included in the HEINLEIN sample Uncheck this box to have NAs inserted instead The advantage of inserting actuals is that it sometimes makes for a pret tier plot of the forecast values The advantage of inserting NAs is that you won t accidentally think you forecasted the values outside HEINLEIN Facing the Unknown So far we ve forecast a number
287. nge and Sample are telling us that we have 28 observations Later we ll see how to change the num ber of observations in the workfile by double clicking on Range and how to change the sample by double clicking on Sample Display Filter is used to control which objects are displayed in the workfile window Dis play Filter is useful if you have hundreds of objects Otherwise it s safely ignored Let s move to the lower panel Our brand new workfile comes with two objects preloaded and The series RESID is designated specially to hold the residuals from the last regression or other statistical estimation See Chapter 3 Getting the Most from Least Squares for a discussion of residuals Since we have not yet run an estimation procedure the RESID series is empty i e all values are set to NA An EViews workfile holds a collection of objects each kind of object designated by its own icon Far and away the most important object is the series icon because that s where our data are stored You ll have noted that the object C has a different icon a Greek letter Instead of a data series C holds values of coefficients Right now C is filled with zeros but if you ran a regression and then double clicked on C you would find it had been filled with estimated coefficients from the last regression Time to Type One Series at a Time We sit with an empty workfile How to bring in the data The easiest way to bring data in
288. nge field in the workfile window is now marked indexed Now if we look at a spreadsheet view of SALARY the rows are labeled with DISCIPLINE in place of an uninformative obser vation number EViews Illustrated book Page 33 Monday February 25 2013 10 06 AM 34 Chapter 2 EViews Meet Data Dated Series Let s set aside our academic salary example for a bit and talk about more options for the identification series and the parallel options for structuring a workfile EViews comes with a rich built in knowledge of the calendar Lots of data lots and lots of data is dated at regular intervals Observations are taken annually quarterly monthly etc EViews understands a variety of such frequencies The only difference between creating an undated workfile and a dated workfile is that for an undated workfile you enter the total number of observations while for a dated workfile you provide a beginning date and an ending date Hint If you want to be able to see more or less of the id series in the left hand column just grab the col umn divider and drag it over to the right All column widths are adjust able in the same way Counting Hint If you want to add the observation number to the label right click and select ObsID To return to the original display just do it again Hint Even when data are measured at regular intervals measurements are sometimes missing Not a problem just leave missing measurements ma
289. nges in the data while a green icon indi cates that updating is off Frozen graphs have two big advantages Customizations are stored as part of the graph object so they don t disappear You can choose whether or not you want the graph to change every time the data or sample changes A good rule of thumb is that if you want any changes to a graph to last freeze it Hint Because a frozen graph with updating off is severed from the underlying data series the options for changing from one type of graph to another categorical graph distribution plot etc are limited It s generally best to choose a graph type before freezing a graph if you intend to keep updating off Hint To make a copy of a graph object perhaps so you can try out new customiza tions without messing up the existing graph click the button and choose Copy Object or press the button to create a graph object with updating turned off A new untitled graph window will open EViews Illustrated book Page 121 Monday February 25 2013 10 06 AM 122 Chapter 5 Picture This A Little Light Customization To add the title US Treasury Interest Rates click the button Enter the title in the Text for label field and change Position to Top The graph now looks almost like the picture opening the chapter The remaining difference is that all the series in this picture are drawn with solid lines while the opening picture used a variety of
290. nometrics visualization alert This average yield curve worked out neatly because the series in the group just happened to be ordered from short maturity to long maturity If we d chosen a different order for the series the line connecting the means wouldn t have been meaningful As is the scaling on the horizontal axis is a lit tle misleading We probably think of the one year rate as being close to the 3 month rate not halfway between the 3 month and 20 year rates EViews Illustrated book Page 130 Monday February 25 2013 10 06 AM Picture One Series 131 Picture One Series Our soup to nuts example graphed three interest rates together Now we step back and for the sake of simplicity look at the various graphic views available for a single series all of which are available by opening a series window and choosing View Graph All these graph types are available for Groups as well as are additional types discussed in Group Graphics below Line Graph and Dot Plots A series line graph is just like the group line graph we saw above except it only shows a single series The line graph plots the value of the series on the vertical axis against the date on the hori zontal axis EViews Illustrated book Page 131 Monday February 25 2013 10 06 AM 132 Chapter 5 Picture This An EViews Dot Plot is a line graph with the lines replaced with little circles Series in a dot plot are indented a little to improve the
291. notice in the dialog that EViews defaults to Dated regular frequency and Annual However the data shown in Table 1 Academic Salary Data Excerpt are just num bered sequentially They aren t dated Choosing the Workfile structure type dropdown menu offers three choices Our data are Unstructured Undated Select this option Later in this chapter we ll discuss Dated regular frequency Balanced Panel is deferred to Chapter 11 Panel What s My Line Hint Alternatively you can type wfcreate in the command pane to bring up the same dialog Hint Changing the type of workfile structure can be mildly inconvenient so it pays to think a little about this decision In contrast simply increasing or decreasing the range of observations in the workfile is quite easy EViews Illustrated book Page 25 Monday February 25 2013 10 06 AM 26 Chapter 2 EViews Meet Data Unstructured Undated instructs EViews to number the observations from 1 through however many obser vations you have In our example we have 28 observations Enter 28 in the field marked Observations If you d like to give your workfile a name you can enter the name in the WF field at the lower right You can also name the workfile when you save it so giving a name now or later is purely a matter of personal prefer ence Hit and the workfile is cre ated for you Deconstructing the Workfile There s no da
292. nowing about this dialog is that it uses series as a special key word You can see an example in the dialog on page 162 When a graph has multiple series EViews will treat data coming from one series versus another analogously with data within a series coming from one category versus another In other words the list of series can be treated like an artificial categorical variable for arranging graph layouts Just as we saw in Multiple Series as Factors The keyword series is used to identify the list of series in the graph Togetherness of the Second Sort At this point you know lots of ways to create graphs Graphs however created are easily combined into a new single graph Hint Remember that a graph view of a series or group isn t a graph object A graph object distinguished by the icon is most commonly created by freezing a graph view EViews Illustrated book Page 164 Monday February 25 2013 10 06 AM Togetherness of the Second Sort 165 Select the graph objects you want to combine just as you would select series for a group or use the show command as in show tm3_ty01 tm3_ty20 where tm3_ty01 tm3_ty20 are names of graphs stored in the workfile to open a new graph object Using the mouse you can re arrange the position of the subgraphs within the overall graph window This allows you to produce some very interesting effects For example if you have two graphs with identical axes you can superimpose one on the o
293. ns the chapter the autocorrelations equal 1 et et 1 et et 2 ut rut 1 e t r r2 r3 EViews Illustrated book Page 318 Monday February 25 2013 10 06 AM Testing for Serial Correlation 319 To plot the autocorrelations of the residuals click the button and choose the menu Residual Diagnostics Correlogram Q Sta tistics Choose the number of autocorrelations you want to see the default 36 is fine and EViews pops up with a combined graphical and numeric look at the autocorrelations The unlabeled column in the middle of the display gives the lag number 1 2 3 and so on The column marked AC gives estimated autocorrelations at the corresponding lag This correlogram shows substantial and persistent autocorrelation The left most column gives the autocorrelations as a bar graph The graph is a little easier to read if you rotate your head 90 degrees to put the autocorrelations on the vertical axis and the lags on the hor izontal giving a picture something like the one to the right showing slowing declining autocorrelations Testing for Serial Correlation Visual checks provide a great deal of information but you ll probably want to follow up with one or more formal statistical tests for serial correlation EViews provides three test sta tistics the Durbin Watson the Breusch Godfrey and the Ljung Box Q statistic Durbin Watson Statistic The Durbin Watson o
294. nt 384 exogenous variables 372 377 expand function 113 expanding data 264 265 explanatory variables 226 227 228 exponential growth 8 exporting to file 243 398 generally 199 graphs 125 126 194 396 412 413 pooled data 312 tables 396 F factor and graph layout options 162 factors 158 multiple series as 161 FALSE 86 87 far outliers 209 fill areas graph 188 192 financial price data conversion of 250 first differences 330 331 fit command 229 fit lines 147 multiple 149 fit of regression line value 66 fit options global 148 fitted values 315 316 EViews Illustrated book Page 417 Monday February 25 2013 10 06 AM 418 Index fixed effects 275 276 282 283 287 289 295 296 308 fixed width text files importing from 46 47 fonts selecting default 408 foo etymology 400 forecast command 229 forecast graphs 146 forecasting ARIMA 332 333 ARMA errors and 327 329 confidence intervals 231 235 236 dynamic 228 229 328 329 example 20 21 223 225 explanatory variables 226 227 in sample 227 logit 351 out of sample 227 rolling 388 serial correlation 327 329 static 228 229 328 329 steps 225 226 structural 328 329 transformed variables 235 238 uncertainty 231 235 236 from vector autoregressions 378 379 verifying 229 231 freezing graphs 119 121 objects 393 394 samples 99 frequency conversion 244 246 251 408 frml command 94 95 103
295. nt to the choice of template will use the template settings Apply template settings to existing uses the template styles for both existing and future text line and shades Replace text amp line shade wipes out existing text line and shade objects and then copies these objects in from the template In the previous figure we manually re entered the zero line after the recession shading was applied from the template This was necessary because the template didn t have a zero line and all lines get replaced Axis and legends in templates Two checkboxes handle how axis and legend settings are copied Check the first box to apply axis settings from the template related to the label and scale The second checkbox will replace the current leg end text with that of the template If neither is checked these items will not be copied Hint Applying a template can make a lot of changes and there s no undo once you exit the dialog It can pay to use Object Copy Object to duplicate the graph before making changes Then try out the template on the fresh copy Hint If you regularly shade graphs to show periods of recession something com monly done in macroeconomics make a template with recession periods shaded and then use Replace text amp line shade to copy the shaded areas onto new graphs Hint Nope there s no way to copy text line or shade objects from the template with out replacing the existing ones EViews Illus
296. ntly The model solver created new series containing the solution values To distinguish these series from our original data EViews adds _0 to the end of the name for each solved Wow hint The model is nonlinear There is no closed form solution This doesn t bother EViews in the slightest Admittedly some models are harder for a computer to solve EViews Illustrated book Page 368 Monday February 25 2013 10 06 AM Your First Homework Bam Taken Up A Notch 369 series That s the source of the series CONS_0 and Y_0 that now appears in the workfile window Open a group in the usual way to look at the solutions The solution for GDP is close to the real data although it isn t perfect Making A Better Model If we d like the solution to be closer to the real data we need a better model In the example at hand we know that we ve left exports and imports out of the model Switch to the text view and edit the identity so it looks as shown here Click again Hint EViews calls a series with the added suffix an alias Hint Discrepancy What discrepancy Well for our data consumption investment government spending and net exports don t quite add up to GDP Welcome to the real world EViews Illustrated book Page 369 Monday February 25 2013 10 06 AM 370 Chapter 15 Super Models Looking At Model Solutions Since EViews placed the results CONS_0 and Y_0 in the work file we can examine
297. nto another program The former copies the internal representation of the graph which can only be pasted into an EViews workfile The picture you want won t show up on the clipboard EViews Illustrated book Page 125 Monday February 25 2013 10 06 AM 126 Chapter 5 Picture This Graph Save To Disk The alternative to copy and paste is to save a graph as a disk file Choose Save to disk either from the button or the right click menu to bring up the Graphics File Save dialog From here you can choose a file format including EMF EPS GIF JPEG PNG PDF and BMP whether or not to use color and whether or not to make the background of the graph transparent You can also adjust the picture size and of course pick a location on the disk to save the file A Graphic Description of the Creative Process Graph creation involves four basic choices What specific graph type should be used to display information Line graph Scatter plot Something more esoteric perhaps Do you want to graph your raw data or are you looking to graph summary statistics such as mean or standard deviation Do want a basic or a categorical graph the latter graph type displaying your data with observations split up into categories specified by one or more control variables For example you might compare wage and salary data for unionized and non union ized workers If more than one group of data is being graphed
298. number crunching is a breeze with EViews modeling facility To get started with making an EViews model use Object New Object to generate a model named KEYNESCROSS The new model object opens to an empty window as shown We re going to type the first equation in manually so hit the button Type in the national income accounting identity When you re done the win dow should look something like the picture shown to the right Hint Since C is a reserved name in EViews we ve substituted CONS for C Hint In this example we typed in one equation and copied another from an estimated equation in the workfile You re free to mix and match although in real work most equations are estimated In addition to linking in an equation object you can also link in SYS and VAR objects Yt ln Ct 1 EViews Illustrated book Page 366 Monday February 25 2013 10 06 AM Your First Homework Bam Taken Up A Notch 367 Let s find out whether we and EViews have had a meeting of the minds on how to interpret the model Click to switch to the equations view which tells us what EViews is thinking You may get a warning mes sage about recompiling the model It can be ignored One line appears for each equation So far there is only one equation Double click on an equa tion for more information for example the Proper ties of the first equation are shown to the right Since this is the national in
299. o check how well your forecasting model works you want to compare forecasts with what actually happens One option is to wait until the future arrives and see how things turned out But the standard procedure is to simulate data arrival by dividing your data sample into an artificial history and an artificial future Our monthly currency data runs from 1917 through 2005 We ll call the complete sample WHOLERANGE treat most of the period as history HERODOTUS and reserve the last few years for a future history HEINLEIN If an equation estimated over HERODOTUS does a good job of forecasting HEINLEIN then we Mea very slightly culpa If you ve been reading really really closely you may have noticed that the forecast for November 2004 in the graph shown in Just Push the Fore cast Button doesn t match the forecast in the table in Theory of Forecasting The former was a dynamic forecast because that s the EViews default and the latter was a static forecast because that s easier to explain Hint The Structural ignore ARMA option isn t relevant in fact is grayed out unless your equation has ARMA errors See Forecasting in Chapter 13 Serial Correlation Friend or Foe EViews Illustrated book Page 229 Monday February 25 2013 10 06 AM 230 Chapter 8 Forecasting can have some confidence that we can re estimate over WHOLERANGE and then forecast out into the yet unseen real future sample wholeRa
300. o omit one series from the complete set expand can take an optional last argument dropfirst or droplast The former omits the first category from the set of series generated and the latter omits the last category For more detail see the Command and Programming Reference expand can also be used in algebraic expressions with each resulting temporary series being inserted in the expression in turn The command ls lnwage c ed expand fe is equivalent to ls lnwage c ed fe 0 ed fe 1 EViews Illustrated book Page 113 Monday February 25 2013 10 06 AM 114 Chapter 4 Data The Transformational Experience and estimates separate returns to education for men and women Statistical Functions Uniform and standard normal random number generators were described earlier in the chapter EViews supplies families of statistical functions organized according to specific probability distributions A function name beginning with r is a random number gener ator a name beginning with d evaluates the probability density function also called the pdf a name beginning with c evaluates the cumulative distribution function or cdf and a name beginning with q gives the quantile or inverse cdf In each case the sign and initial letter are followed by the name of the distribution As an example the name used for the uniform distribution is unif So runif a b gen erates random numbers dis
301. odel logit Logit Instead of fitting zeros and ones the logit model uses the right hand side variables to predict the probability of being a union member i e of observing a 1 0 One can think of the model as having two parts First an index is created which is a weighted combination of the explanatory variables Then the probability of observing the outcome depends on the cumu lative distribution function cdf of the index Logit uses the cdf of the logistic distribution probit uses the normal distribution instead The logit command is straightfor ward logit union c age The coefficients shown in the output are the coefficients for constructing the index In this case our estimated model says Textbook hint Textbooks usually describe the relation between probability and index in a logit with rather than The two are equivalent for a logit or a probit but differ for some other models s s 4 23 0 028 age prob union 1 1 F s F s es 1 es prob union 1 F s prob union 1 1 F s EViews Illustrated book Page 350 Monday February 25 2013 10 06 AM ARCH etc 351 Logit s Forecast dialog offers a choice of predicting the index s or the proba bility Here we predict the probability The graph shown to the right plots the probability of union member ship as a function of age For c
302. of observations used in the table Table Interpretation In this applicant pool there were 5 stu dents with perfect grades and below aver age LSATs That s a true fact but so what We might be interested in getting counts but usually what we re trying to do is find out if one variable is related to another For the data in hand the obvious question is Do high test scores and high grades go together To begin to answer this question we return to the N Way Tabulation menu this time checking Table Row and Column in the Crosstabulation dialog Hint The intersection of a row and column is called a cell For example there are 350 applicants in the High grade Low test score cell EViews Illustrated book Page 217 Monday February 25 2013 10 06 AM 218 Chapter 7 Look At Your Data Take a look at the Top grade Low test score cell again The first num ber in the cell tells us as before that five applicants had this combination of grade and LSAT But now we have three additional numbers The first new number marked is the Table 0 31 5 1638 which tells us the fraction of observations falling in this cell out of all the observations in the table The Row tells us what fraction of top grades the row also have low test scores 5 10 Analogously the last element in the cell is the Col umn 5 775 In the same way table row and column percentage are given in the Total column
303. om parison purposes we ve added a horizontal line marking the uncon ditional probability of union mem bership A 60 year old is about three times as likely to be in a union as is a 20 year old ARCH etc Have you tried About EViews on the Help menu and then clicked the button Only one Nobel prize winner so far appears in the credits list Which brings us to the topic of autoregressive conditional heteroskedasticity or ARCH ARCH and members of the extended ARCH family model time varying variances of the error term The simplest ARCH model is In this ARCH 1 model the variance of this period s error term depends on the squared residual from the previous period yt a b xt et jt 2 g0 g1e2 t 1 EViews Illustrated book Page 351 Monday February 25 2013 10 06 AM 352 Chapter 14 A Taste of Advanced Estimation The residuals from the cur rency data used earlier showed noticeably persistent volatility a sign of a potential ARCH effect In EViews all the action in specifying ARCH takes place in the Specifica tion tab of the Equation Esti mation dialog To get to the right version of the Specifica tion tab choose ARCH Autoregressive Conditional Heteroskedasticity in the Method dropdown of the Esti mation settings field Hint Unlike nearly all other EViews estimation procedures ARCH requires a continu ous sample Define an appropriate sample in the Specification tab If yo
304. ometimes its better to break links to avoid such unintended changes Quick Review A page is fundamentally a workfile within a workfile You can use multiple pages simply as a convenient way to store different sets of data together in one workfile The real power of pages lies in the fact that each page can have a different identifier Series can be brought from one page to another either by copying the values in the source page into the destination page or by creating a live link If you create a link EViews will fetch a fresh copy of the data every time the link series is referenced Not only will EViews copy data it will also translate data from one identifier to another Because EViews is big on calendars it has a bag of tricks for converting one frequency to another x xi i 1 n n EViews Illustrated book Page 267 Monday February 25 2013 10 06 AM 268 Chapter 9 Page After Page After Page Even where the identifier is something other than time you can contract data by supplying a rule for selecting sets of observations and then summarizing them in a single number For example you might contract data on individuals by taking state wide averages Inversely you can also expand data by instructing EViews to look up the desired values in a table EViews Illustrated book Page 268 Monday February 25 2013 10 06 AM Chapter 10 Prelude to Panel and Pool So far our data has come in a simple peaceful arrangem
305. on 249 250 Drag and drop copy 246 new page 241 243 dummies including manually 287 289 dummy variables 99 Durbin Watson DW statistic 73 281 319 320 324 dynamic forecasting 228 229 328 329 E Econometric Theory and Methods Davidson amp MacKinnon 320 EMF 126 EMF enhanced metafile files 125 empirical distribution tests 207 encapsulated postscript EPS files 125 endogenous variables 372 377 enhanced metafile EMF format files 125 EPS 126 EPS encapsulated postscript files 125 EQNA function 92 equal function 84 87 105 error bar graphs 145 146 errors ARMA 229 323 324 329 data 197 198 estimated standard 373 moving average 322 323 324 325 standard 66 69 errors in variables 346 Escape Esc key 401 estimated standard error 373 estimation 2SLS 345 347 ARCH 351 355 GMM 347 348 heteroskedasticity 338 340 limited dependent variables 349 351 maximum likelihood 355 357 nonlinear least squares 341 345 of standard errors 308 options 411 ordinary least squares 359 360 panels 279 283 284 285 287 289 pools 294 304 308 random 308 regression 317 speed 393 system 357 361 weighted least squares 335 338 evaluated strings 383 EViews automatic updates 400 customizing 404 409 data capacity 393 data limitations 393 help resources 401 speed 393 updates fixes 400 401 EViews Forum 1 EViews ini 412 Excel importing data from 40 43 exclamation poi
306. on Axis borders We can also add confidence ellipses through the Fit lines menu and the button Here we ve added histograms to the axes in our law school admission exam ple as well as two confidence ellipses The confidence ellipses enclose areas that would contain 90 percent and 95 percent respectively of the sample if the sample was drawn from a normal distribution Of course as you can see from the histograms the data are not drawn from a normal distribution EViews Illustrated book Page 151 Monday February 25 2013 10 06 AM 152 Chapter 5 Picture This XY Line and XY Area Graphs XY Line graphs are really just scatter plots where the consecu tive points are connected and XY Area graphs are XY Line graphs with the area below the line filled in Contrast a scatter diagram shown right of infla tion versus unemployment Output_and_Unemployment w f1 from 1959 through 1979 with the same data shown with connecting lines shown next The connecting lines give a much clearer hint that we re see ing a series of negatively sloped relatively flat lines that are mov ing up over time Let s try a little trick for display ing pre 1970 and 1970 s inflation separately on the same graph By making separate series for inflation in different periods we can exploit the ability of XY Line graphs to show multiple pairs We create our series with the commands smpl 59 69 series inf_early inflation sm
307. on Effects view The reported values of the cross section fixed effects are the intercept for country i less the average intercept So it s not very surprising that the effect for Canada is positive and the effect for the Central African Republic is negative Hint Fixed Random Effects Testing offers a formal test for the presence of fixed effects Look for it on the View menu Hint When using fixed effects the constant term reported in regression output is the average value of ai ai EViews Illustrated book Page 282 Monday February 25 2013 10 06 AM Pretty Panel Pictures 283 Since the results differ dramati cally it would be nice to have some assurance that the fixed effects are really there Panel estimates include extra coeffi cient testing views Choose Fixed Random Effects Test ing Redundant Fixed Effects Likelihood Ratio In this case the statistical evidence as shown by the p value is over whelmingly in favor of keeping fixed effects in the model Pretty Panel Pictures EViews offers extra ways of looking at the panel structured data especially in graphs When you select the Graph view for data in a panel struc tured workfile the dialog offers additional panel options in the bottom right corner Choosing Combined cross sec tions gives the figure to the right Mean anything to you Me neither There are just too many darn lines in this picture EViews Illustrated book Page 283 Monday
308. on both X1 and X2 would use only observations 1 and 4 The variable X1 1 giving the previous period s values of X1 is missing both the first and third observation The first value of X1 1 is NA because the data from the obser vation before observation 1 doesn t exist There is no observation before the first one eh The third observation is NA because it s the second observation for X1 and that one is NA So while a regression of Y on X1 would use observations 1 3 4 and 5 a regression of Y on X1 1 would use observations 2 4 and 5 Moral When there s missing data changing the variables specified in a regression can also inadvertently change the sample EViews Illustrated book Page 70 Monday February 25 2013 10 06 AM The Pretty Important But Not So Important As the Last Section s Regression Results 71 The third line just reports the date and time EViews estimated the regression It s surprising how handy that information can be a couple of months into a project when you ve forgot ten in what order you were doing things Since we re talking about looking at output at a later date this is a good time to digress on ways to save output for later You can Hit the button to save the equation in the workfile The equation will appear in the workfile window marked with the icon Then save the workfile Hit the button Spend output to a Rich Text For mat RTF file which can then be r
309. on or expression you can reference the series in another page by name using the form workfile_page_name series_name For example say you have a workfile with two pages MYPAGE1 and MYPAGE2 where the first page has a series Y and the second page has a series X If you want the values of Y to be log X you can simply type Y log mypage2 X in the command pane EViews will take the log values of X from MYPAGE2 and initialize Y in MYPAGE1 accordingly And you haven t cluttered your workfile with an extra link series If frequency conversion between the pages is necessary default frequency conversion will be used Unlinking Suppose that the data in your source page is regularly updated but you want to analyze a snapshot taken at a point in time in the desti nation page To freeze the values being linked in open the link and choose or and Unlink Alter natively with the workfile window active choose Object Manage Links amp Formulae The dialog lets you manage links and for mulae in your workfile The Break Links convert into ordinary series button detaches the specified links from their source data and converts them into regular numeric or alpha series using the current values in all links or just those you list Hint Of course you can use the series command series frozen_copy_of_series linked_series to make a copy of a linked series as an alternative to breaking the link EViews Illustrated book
310. on tells EViews to squish the data before storing to disk When much of the data takes only the values 0 or 1 which is quite common disk file size can be reduced by nearly a factor of 64 This level of compression is unusual but shrinkage of 90 percent hap pens regularly Unlike use of single precision compression does not cause any loss of accu racy The truth is that disk storage is so cheap that there s no reason to try to conserve it While disk storage itself is rarely a limiting factor moving around large files is sometimes a nui sance This is especially true if you need to email a workfile Hint Workfile Save Options have no effect on the amount of RAM internal storage required They re just for disk storage EViews Illustrated book Page 410 Monday February 25 2013 10 06 AM Estimation Defaults 411 Earlier releases of EViews before 5 0 can read single precision but not compressed files Estimation Defaults The Estimation options dialog lets you set defaults for control ling the iteration pro cess and internal computation of deriva tives in nonlinear esti mation There s nothing wrong with the out of the box defaults although some people do prefer a smaller number for the Conver gence value You can also set these controls as needed for a specific estimation problem but if you do lots of nonlinear estimation you may find it convenient to reset the defaults here Hint There s no reason not
311. ook at our data this time as displayed in the panel workfile Now the obs column correctly identifies each data point with both country code and year That s about all you need to know to set up a panel workfile One more option is worth mentioning Should you balance A panel is said to be balanced when every cross section is observed for the same time period The Penn World Table data are bal anced since there are observations for 1950 through 2000 for every country although quite a few observations are simply marked NA not available If you look at the workfile you ll see that all the data for the Central African Republic is missing until 1960 The Cen tral African Republic became independent on August 13 1960 The creators of the Penn World Table might have simply omitted these years for the Central African Republic giving us an unbalanced panel EViews default if you leave the check box Balance between starts amp ends in the Workfile structure dialog checked is to make a balanced workfile by insert ing empty rows of data where needed Use Balance between starts amp ends unless you have a reason not to do so See the User s Guide for further discussion Panel Estimation Not that it was much trouble but we didn t restructure the workfile just to get a prettier dis play for spreadsheet views Let s work through an estimation example EViews Illustrated book Page 278 Monday February 25 2013 10 06 AM
312. or extreme accuracy no satis factory answer can be found Accurately understanding accuracy Don t confuse numerical accuracy with model accuracy The solver options control numerical accuracy These options have nothing to do with the accuracy of your model or your data The latter two are far more impor tant Unfortunately you can t improve model or data accuracy by clicking on a button EViews Illustrated book Page 374 Monday February 25 2013 10 06 AM Your Second Homework 375 Making Scenarios EViews puts a single set of assumptions about the inputs to a model together with the resulting solution in a scenario The solu tions based on the original data are called the Baseline So the solutions to our first homework problem are stored in the baseline scenario Choosing Scenarios from the View menu brings up the Scenario Specification dialog with the Select Scenario tab showing The Baseline scenario that s show ing was automatically created when we solved the model A look at the aliasing tab shows that the suffix for Baseline results is _0 The fields are greyed out because EViews assigns the suffix for the Baseline EViews Illustrated book Page 375 Monday February 25 2013 10 06 AM 376 Chapter 15 Super Models We want to ask what would hap pen in a world in which govern ment spending were 10 billion dollars higher than it was in the real world This is a new scenario so click on the Sel
313. or you to care about On the author s somewhat antiquated PC computing a linear regression with ten right hand side variables and 100 000 observations takes roughly one eye blink Nonlinear estimation can take longer First a single nonlinear estimation step i e one iter ation can be the equivalent of computing hundreds of regressions Second there s no limit on how many iterations may be required for a nonlinear search EViews nonlinear algo rithms are both fast and accurate but hard problems can take a while Freeze EViews objects change as you edit data change samples reset options etc When you have table output that you want to make sure won t change click the button to open a new window disconnected from the object you ve been working on and therefore frozen You ve taken a snapshot Nothing prevents you from editing the snapshot that s the stan Daughter hint Oh daddy that s so 90 s Hint Student versions of EViews place limits on the amount of data that may be saved and omit some of EViews more advanced features EViews Illustrated book Page 393 Monday February 25 2013 10 06 AM 394 Chapter 17 Odds and Ends dard approach for customizing a table for example but the frozen object won t change unless you change it Frozen graphs offer more sophisticated behavior than frozen tables When you have graphical output and click on the but ton EViews opens a dialog prompting yo
314. oregressive and moving average errors can be combined into an autoregressive moving average or ARMA process For example putting an AR 2 together with an MA 1 gives you an ARMA 2 1 process which can be written as For an ARMA 3 4 model you would traditionally have to include a separate term for each ARMA lag such that you would write it in your equation as ls y c ar 1 ar 2 ar 3 ma 1 ma 2 ma 3 ma 4 In EViews you can just write ls y c ar 1 to 3 ma 1 to 4 It works in the same way for lags you can write y 1 to 4 instead of having to write y 1 y 2 y 3 y 4 Correcting for Serial Correlation Now that we know that our stock volume equation has serial correlation how do we fix the problem EViews has built in features to correct for either autoregressive or moving average errors or both of any specified order The corrected estimate is a member of the class called Generalized Least Squares or GLS For example to correct for first order serial corre lation include AR 1 in the regression command just as if it were another variable The command ls logvol c trend trend 2 d log close 1 ar 1 gives the results shown The first thing to note is the additional line reported in the middle panel of the output The serial correlation coeffi cient what we ve called in writing out the equations is labeled AR 1 and is estimated to equal about 0 84 The associated standard error t sta
315. other words not worrying whether observa tions are missing for series two etc Here we see the correlation for the two series in the group Surprisingly LSAT and GPA are not all that highly correlated A correlation of only 0 307 means that the information in LSAT is not redun dant with the information in GPA which is why law schools look at both test scores and grades People use correlations a lot more because correla tions are unit free while the units of covariances depend on the units of the underlying series EViews Illustrated book Page 215 Monday February 25 2013 10 06 AM 216 Chapter 7 Look At Your Data Cross Tabs Cross tabulation is a traditional method of looking at the relationship between categorical variables In the simplest version we build a table in which rows represent one variable and columns another and then we count how many observations fall into each box For fun we ve created categorical variables describing grades 4 0 or better above average but not 4 0 below average and test scores above average below average with the following com mands series gradecat gpa gt 4 1 gpa gt 3 365 and gpa lt 4 2 gpa lt 3 365 3 series testcat lsat gt 158 1 lsat lt 158 2 Since GRADECAT is arbitrarily coded as 1 2 or 3 and TESTCAT is similarly arbitrarily coded as 1 or 2 we added a value map to each series to make the tables easier to read See What Are Your Values in Chapter
316. our solution using any of the usual tools for looking at series In addition the model object has tools con venient for this task in the Proc menu Choose Proc Make Graph to bring up the Make Graph dia log and click The graph window that opens shows the time path for all the variables in our homework model EViews Illustrated book Page 370 Monday February 25 2013 10 06 AM Looking At Model Solutions 371 If you prefer to see all the series on a single graph choose the radio button Group by Scenario Actu als Deviations etc in the Make Graph dialog To make the graph prettier we ve reassigned Y and CONS to the right axis See Left and Right Axes in Group Line Graphs in Chapter 5 Pic ture This Comparing Actual Data to the Model Solution Return again to the Make Graph dialog This time choose Listed variables enter Y in the text field and check the checkbox Com pare and choose Actuals on the dropdown menu Click Group by Model Variable so that all the GDP data will appear on the same graph If you d like also check Deviation Active from Compare EViews Illustrated book Page 371 Monday February 25 2013 10 06 AM 372 Chapter 15 Super Models You can see that we now do a much better job of matching the real data More Model Information Before we see what else we can do with a model let s explore a bit to see what else is stored inside Model Variables Click the bu
317. own to the right You already knew some of the numbers in this regression because they appeared in the sec ond column in Table 1 on page 65 When you specify a multiple regression EViews gives one row in the output for each indepen dent variable Hypothesis Testing We ve already seen how to test that a single coefficient equals zero Just use the reported t statistic For example the t statistic for lagged log volume is 37 89 with 460 degrees of free dom 464 observations minus 4 estimated coefficients With EViews it s nearly as easy to test much more complex hypotheses Hint Most regression specifications include an intercept Be sure to include C in the list of independent variables unless you re sure you don t want an intercept Hint Did you notice that EViews reports one fewer observation in this regression than in the last and that EViews changed the first date in the sample from the first to the second quarter of 1888 This is because the first data we can use for lagged volume from second quarter 1888 is the non lagged volume value from the first quarter We can t compute lagged volume in the first quarter because that would require data from the last quarter of 1887 which is before the beginning of our workfile range EViews Illustrated book Page 74 Monday February 25 2013 10 06 AM Hypothesis Testing 75 Click the button and choose Coefficient Diag nostics Wald Coefficient Restrictions
318. ows average academic salaries and corresponding salaries outside of academics Table 1 Academic Salary Data Excerpt OBS DISCIPLINE SALARY NONACADSAL 1 Dentistry 44 214 40 005 2 Medicine 43 160 50 005 3 Law 40 670 30 518 4 Agriculture 36 879 31 063 5 Engineering 35 694 35 133 6 Geology 33 206 33 602 7 Chemistry 33 069 32 489 8 Physics 32 925 33 434 9 Life Sciences 32 605 30 500 10 Economics 32 179 37 052 28 Library Science 23 658 15 980 EViews Illustrated book Page 23 Monday February 25 2013 10 06 AM 24 Chapter 2 EViews Meet Data The Structure of Data and the Structure of a Workfile Look at Table 1 Academic Salary Data Excerpt First thing to notice data come arranged in rows and columns Every column holds one series of data for example the values of SALARY for every discipline Every row holds one observation an example being the value of SALARY NONACADSAL and the name of the discipline for dentistry When data come arranged in a neat rectangle as it does here stat isticians call the arrangement a data rectangle Second thing to notice the observations rows come in order In the column marked obs the observations are numbered 1 2 3 4 28 The observation numbers are sometimes called well observation numbers Sometimes the entire set of observation numbers is called an identifier or an id series
319. p in columns The data in academic salaries by discipline txt lines up pretty much the same way as did the same data in the Excel file we looked at above To pull text data directly into an workfile use File Open Foreign Data as Workfile and point to the appropriate text file You don t want File Open Text File that s for bringing a file in as text not for converting the text to an EViews workfile EViews pops up with the ASCII Read dia log EViews has analyzed our text file and made a judgment call about how to interpret the data A quick glance shows that EViews has hit it spot on so we can just hit and we ll have our workfile Text files aren t always this easy to interpret When EViews reads in a line it has to decide which information goes with which variable In this example data are separated by tabs When EViews finds a tab it knows it s done reading the current datum In this context a tab is called a delimiter because it marks de limit or de boundary of a column EViews has a built in facility for using tabs or spaces for delimiters and also allows you to customize the EViews Illustrated book Page 44 Monday February 25 2013 10 06 AM The Import Business 45 choice of delimiter These choices are found by hitting in the ASCII Read dia log Space delimited The most common format for text data is probably space delimited That just means there are one or more spaces
320. pl 70 79 series inf_late inflation smpl 59 79 The series INF_EARLY has inflation from 1959 through 1969 and NAs elsewhere Similarly INF_LATE is NA except for 1970 through 1979 Now we open a group with the unemploy ment rate as the first series and INF_EARLY and INF_LATE as the second and third series We can exploit the fact that EViews doesn t plot observations with NA values to get a very nice looking XY Line graph EViews Illustrated book Page 152 Monday February 25 2013 10 06 AM Group Graphics 153 Pie Graphs Pie graphs don t fit neatly into the EViews model of treating a series as the relevant object The Pie Graph command produces one pie for each observation The observation value for each series is converted into one slice of the pie with the size of the slice representing the observation value of one series relative to the same period s observation for the other series For example if there are three series with values the first two slices will each take up one quarter of the pie and the third slice will occupy the remaining half Obscure graph type hint The graph type XY Bar X X Y triplets lets you draw bar graphs semi manually Vertical bars are drawn with the left edge specified by the first series the right edge specified by the second series and the height given by the third series p p and 2p EViews Illustrated book Page 153 Monday February 25 2013 10 06 AM 154 Chapter 5 Picture Th
321. pre ceding observation which may or may not have been measured one time period ear lier Put another way in a workfile holding data for U S states in alphabetical order one lag of Missouri is Mississippi t 1 EViews Illustrated book Page 88 Monday February 25 2013 10 06 AM Your Basic Elementary Algebra 89 The first assignment statement sets all the observations of Y to 1 As a consequence of the second smpl statement Smpl limits operations to a subset of the data more on smpl in the section Simple Sample Says the second assignment statement begins with the second observation setting Y to the value of the first observation Y 1 plus 5 1 0 0 5 Then the statement sets the third observation of Y to the value of the second observa tion Y 1 plus 5 1 5 0 5 Contrast this with processing the entire series at a time adding 5 to each original lagged obser vation setting all values of Y to 1 5 which is what EViews does not do Now we ll unignore the smpl statements If we d simply typed the commands series y 1 y y 1 5 the first assignment would set all of Y to 1 But the second statement would begin by adding the value of the zeroth obser vation Y 1 oops what zeroth observation Since there is no zeroth observation EViews would add NA to 0 5 setting the value of the first Y to NA Next EViews would add the first observation to 0 5 this time setting the second Y to NA 0 5
322. r shows that follow ing a shock to the G equation G wig gles around for a quarter or so but by the fifth quarter the response has effectively dissi pated In contrast the upper right hand corner graph shows that G is effectively unresponsive to shocks in the GV equation Variance decomposition How much of the variance in G is explained by shocks in the G equation and how much is explained by shocks in the GV equation The answer depends on among other things the estimated coeffi cients the estimated standard error of each equation and the order in which you evaluate the shocks View Variance Decomposition leads to the VAR Vari ance Decompositions dialog where you can set various options Hint The dashed lines enclose intervals of plus or minus two standard errors EViews Illustrated book Page 363 Monday February 25 2013 10 06 AM 364 Chapter 14 A Taste of Advanced Estimation The variance decomposition shows one graph for the variance of each equation from each source The horizontal axis tells the num ber of periods fol lowing a shock to which the decom position applies and the vertical axis gives the frac tion of variance explained by the shock source In this example most of the vari ance comes from the own shock i e G shocks effect on G rather than from the shock to the other equa tion Forecasting from VARs In order to forecast from a VAR you need to use the model o
323. r 9 Page After Page After Page Contracted Data What we ve done is called a contraction because we ve mapped many data points into one We can see that the unionization rate in Arkansas home of the world s largest private employer is about a half percent and the average education level is three quarters of a year of college In Washington where the state bird is the geo duck the unionization rate is over three percent and average education is about a year and a half of college Something s wrong Unionization rates aren t that low To help inves tigate let s copy the data in using a count merge instead of a mean merge We click on the tab to return to the Cps page re copy the four series and paste into the ByState page as before except with two dif ferences In the Pattern field in the Paste Special dialog we add the suffix count to the variable names so that we don t write over the state means that we computed previously In the Contraction method field switch to Number of obs to get a count of how many observations are being used for each series EViews Illustrated book Page 262 Monday February 25 2013 10 06 AM Contracted Data 263 We can see that education and unionization always have the same underlying counts within a state But the count for LNWAGE is different and lower than the count for ED and UNION Here s what happened in the mean merge The contraction computed the mean
324. r DW statistic is the traditional test for serial corre lation For reasons discussed below the DW is no longer the test statistic preferred by most econometricians Nonetheless it is widely used in practice and performs excellently in most situations The Durbin Watson tradition is so strong that EViews routinely reports it in the lower panel of regression output The Durbin Watson statistic is unusual in that under the null hypothesis no serial correlation EViews Illustrated book Page 319 Monday February 25 2013 10 06 AM 320 Chapter 13 Serial Correlation Friend or Foe the Durbin Watson centers around 2 0 rather than 0 You can roughly translate between the Durbin Watson and the serial correlation coefficient using the formulas If the serial correlation coefficient is zero the Durbin Watson is about 2 As the serial corre lation coefficient heads toward 1 0 the Durbin Watson heads toward 0 To test the hypothesis of no serial correlation compare the reported Durbin Watson to a table of critical values In this example the Durbin Watson of 0 349 clearly rejects the absence of serial correlation The Durbin Watson has a number of shortcomings one of which is that the standard tables include intervals for which the test statistic is inconclusive Econometric Theory and Meth ods by Davidson and MacKinnon says the Durbin Watson statistic despite its popularity is not very satisfactory the DW statistic is
325. r a serious scientific investigation we d have to turn to a data source other than the CPS in order to get a more meaningful socio economic break down EViews Illustrated book Page 286 Monday February 25 2013 10 06 AM Fixed Effects With and Without the Social Contrivance of Panel Structure 287 Our new results estimate that the Asian effect is negative although not significantly so rather than positive Our specu lation that the positive Asian effect was picking up location effects appears to be correct Fixed Effects With and Without the Social Contrivance of Panel Structure EViews provides a large set of features designed for panel data but fixed effects estimation is the most important In terms of econometrics specifying fixed effects in a linear regres sion is a fancy way of including a dummy variable for each group country or state in the examples in this chapter You re free to include these dummy variables manually if you wish There is a special circumstance under which including dummies manually is required Once in a while you may have three or more dimensional panel data Since EViews panels are limited to two dimensions the only way to handle a third dimension of fixed effects is by adding dummy variables in that third dimension by hand There is a fairly common circumstance under which including dummies manually may be preferred If all you re after is fixed effects why bother setting up a panel struc
326. r on the horizontal axes dates some times appear on the vertical rotated graphs are a notable example The appropriate marking options work as you would expect EViews Illustrated book Page 186 Monday February 25 2013 10 06 AM Options Options Options 187 Legend The Legend sec tion controls a number of options the most useful of which is editing Legend entries Gener ally a series Dis play Name you can edit the dis play name from the series label view is used to identify the series in the leg end If the series doesn t have a Display Name the series name itself is used Either way this is the spot for you to edit the legend text On the Attributes page the Legend Columns entry on the left side determines how many columns are used in the legend The default Auto automatic lets EViews use its judg ment Alternatively select 1 2 3 etc to specify the number of columns Legend Aesthetics Setting the text for the legend sometimes presents a trade off between aesthetics and informa tion The longer the text the more information you can cram in But shorter legends generally look better Here s a graph with a moderately long legend Rule of thumb the legend should be shorter than the frame EViews Illustrated book Page 187 Monday February 25 2013 10 06 AM 188 Chapter 6 Intimacy With Graphic Objects Here s the same graph with shorter legend text This graph looks better a
327. r the desired descriptive statistics and EViews will 1 create the requested series YMEAN YMED etc and 2 open an untitled group displaying the new series EViews Illustrated book Page 301 Monday February 25 2013 10 06 AM 302 Chapter 12 Everyone Into the Pool Getting Out of the Pool Pooled series are plain old series so the pool object pro vides a number of tools for manipulating pooled plain old EViews objects in convenient ways A number of useful procedures involving pools appear under the Proc menu Pool Series Generation Much of our analysis on the pool has used percentage population growth measured as D LOG POP We might want to generate this as a new series for each country Manually we could give six commands of the form series dlpcan d log popcan series dlpfra d log popfra To automate the task hit the button and enter the equation using a everywhere you want the country identifier to go The entry DLP D LOG POP generates all six series Hint If you don t see the procedures for pools listed under the Proc menu be sure that the pool window is active Clicking the button in a pool window gets you to the same menu EViews Illustrated book Page 302 Monday February 25 2013 10 06 AM Getting Out of the Pool 303 Pool Series Degeneration To delete a pile of series choose Delete Pool series from the Proc menu This deletes the series It doesn t affe
328. raph Open the graph sans shades and click You may need to widen the graph window to see this button on the far right The Graph Options dialog opens to the Templates amp Objects section I once made a graph with the NBER recessions marked as shades and saved that graph in the workfile with the name RECESSIONS Choos ing RECESSIONS in the Graphs as templates scroll area makes available all the graphic options including shade bands for copying into the current graph Once you hit the button all the basic graphic elements are copied into the current graph with the exception of text objects line and shade objects some axis options and leg end text Hint When you select a template an alert pops open to remind you that all the options in all the various Graph Options sections will be changed Hint The list in Graphs as templates shows graphs in the current workfile If the tem plate graph is in another workfile it s easy enough to copy it and then paste it into the current workfile EViews Illustrated book Page 174 Monday February 25 2013 10 06 AM Templates for Success 175 Text lines and shades in templates Radio buttons offer three options for how text line and shade objects are copied Do not apply means don t copy these objects from the template Use this setting to change everything except text line and shades to the styles in the template Any text line or shade objects you add subseque
329. re independent That s exactly what s done in the lines marked Test Statistics which appear above the table Formally if the two series were independent then the reported test statistics would be approximately with the indicated degrees of freedom The column marked Prob gives p values So despite what we found in the Top grade Low test cell the typical cell percentage is sufficiently different from the product of the marginal percentages that the hypothesis of independence is strongly rejected N Way Tabulation with N gt 2 With two series in a group the cells are laid out in a two dimensional rectangle with the categories for the first series going down and categories for the second series going across With three series EViews displays a three dimensional hyper rectangle With four series the display is a four dimensional hyper rectangle etc Fortunately EViews is very clever at detecting older equipment If you are still using a dis play limited to two dimensions rather than one of the newer Romulan units EViews splits the hyper rectangle into a series of two dimensional slices As an example open a group with GRADECAT TESTCAT and WASH to add the effect of in state residency into the mix Then choose N Way Tabulation Reporting of test statis tics is turned off simply because the output became very long Statistics hint It s not uncommon to find many cells containing very few observations As a rule of th
330. recasts Past that you need to know how to push the but ton EViews Illustrated book Page 238 Monday February 25 2013 10 06 AM Chapter 9 Page After Page After Page An EViews workfile is made up of pages Pages extend our ability to organize and analyze data in powerful ways Some of these extensions are available with nothing more than a sin gle mouse click while others require quite a bit of thought This chapter starts with a look at the easy extensions and gradually works its way through some of the more sophisticated applications As a practical matter most workfiles have only a single page and you ll never even notice that the page is there We think of a workfile as a collection of data series and other objects all stuffed together for easy access In some technical sense the default workfile contains a single page and all the series etc are inside that page Because almost all EViews opera tions work on the active page and because when there is only one page that s the page that s active the page is effectively transparent in a single page workfile In other words you don t have to know about this stuff On the other hand flipping pages can be habit forming Pages let you do a variety of neat stuff like Pull together unrelated data for easy accessibility Easiest Hold multiple frequency data e g both annual and quarterly in a single workfile Easy Link data with differing identifier se
331. rence are just a Help menu click away Quick Help Reference provides quick links to summaries of com mands object properties etc EViews Help Topics leads to extensive indexed and searchable help on nearly everything in EViews plus nice explanations of the underlying econometrics Most of the material in the help system gives the same information found in the manuals but sometimes it s easier and faster to find what you re after in the help sys tem Odd Ending We hope EViews Illustrated has helped but reading is rarely quite so enlightening as doing It s time to click buttons and pull down menus and type stuff in the command pane and generally have fun trying stuff out EViews Illustrated book Page 401 Monday February 25 2013 10 06 AM 402 Chapter 17 Odds and Ends EViews Illustrated book Page 402 Monday February 25 2013 10 06 AM Chapter 18 Optional Ending EViews devotes an entire menu to setting a myriad of Options A little bit of one time customiza tion makes EViews a lot more comfortable and detailed cus tomization can really speed along a big project We ll explore options in three levels of detail starting in this section with what you absolutely have to know The next section gives you our personal recom mendations for changing set tings The final section of the chapter walks through all the other important options Required Options If it s required then it really isn t an option
332. ries into a single analysis Moderate Data reduction Moderate to hard Pages Are Easy To Reach Every EViews workfile has at least one page Page names appear as a series of tabs at the bottom of the workfile window When a workfile is created it con tains a single page usually titled Untitled You can tell that Untitled is the active page not only because it s the only page there is but also because the tab for the active page is displayed with a white background and slightly in front of the other tabs EViews Illustrated book Page 239 Monday February 25 2013 10 06 AM 240 Chapter 9 Page After Page After Page Clicking on a page tab makes that page active What we see in the workfile window is actually the contents of the active page For example choosing the CPS tab in the workfile CPSMar2004 With Pages wf1 displays data on 136 879 individuals Click instead on the ByState tab and we see data on states instead Both sets of data are stored in the same workfile Typically we work with one data set at a time but one of the things we do in this chap ter is show how to link data from one workfile page to another Creating New Pages To add a second page to a workfile click on the New Page tab to bring up a menu with a number of options Three of the menu options represent standard methods that are similar to creating a workfile except that what you actually get is a page within a workfile ins
333. rked NA EViews Illustrated book Page 34 Monday February 25 2013 10 06 AM Dated Series 35 Let s create a couple of workfiles for practice As a first example let s make an annual workfile for the Roosevelt years Franklin not Teddy Use the menu command File New Workfile to bring up the Workfile Create dia log Fill in the fields as shown Note that when the workfile window pops open Range shows 1933 to 1945 and that there are 13 observations in the workfile A second example Most national income accounting macro data for the United States is available on a quar terly basis starting in 1947 To set up a quarterly workfile use File New Workfile and change the drop down menu Date Specifica tion Frequency to Quarterly A new issue arises what are the formats for specifying dates Rules for date for mats are one of those boring but nec essary details that we ll put off til a boring but necessary appendix at the end of the chapter Why add a date structure to a workfile One minor reason is that it saves you the trouble of figuring out that 1947q1 through 2004q4 includes exactly 232 observations There are two more important reasons An understanding of the calendar is built into many operations so it pays to tell EViews how your information is dated Two examples that we ll look at later EViews EViews Illustrated book Page 35 Monday February 25 2013 10 06 AM 36 Chapter 2 EViews Meet D
334. rml We ll meet the latter two commands shortly Genr will still work even though the new commands are pre ferred Computer folks say an old feature has been deprecated when it s been replaced by something new but the old feature continues to work Hint If you aren t sure about the order of operations extra parentheses do no harm EViews Illustrated book Page 86 Monday February 25 2013 10 06 AM Your Basic Elementary Algebra 87 EViews also provides the logical operators and or and not EViews evaluates arithmetic operators first then comparisons and finally logical operators Using 1 and 0 for TRUE and FALSE sets up some incredibly convenient tricks because it means that multiplying a number by TRUE copies the number while multiplying by FALSE gives nothing uh zero isn t really nothing but you know what we meant For example if the series ONE_2_3 and TWO_3_1 are as shown then the command series bigger one_2_3 gt two_3_1 one_2_3 one_2_3 lt two_3_1 two_3_1 picks out the values of ONE_2_3 when ONE_2_3 is larger than TWO_3_1 and the values of TWO_3_1 when TWO_3_1 is larger The Lag Operator Reflecting its time series origins a couple of decades back EViews takes the order of obser vations seriously In standard mathematical notation we typically use subscripts to identify one observation in a vector If the generic label for the current observation is then the previous observ
335. rness of the First Sort Scatter plots XY Line plots too have a special feature by which you can include multiple fitted lines on one scatter plot We ll use this feature to illustrate what happens when you misspecify the functional form in a regression First we ll generate some artificial data workfile u 100 series x rnd series log y 2 3 x rnd Now we ll make a scatter plot and have EViews put in the standard misspecified linear regression line Notice the predominance of positive errors for both low and high values of X EViews Illustrated book Page 149 Monday February 25 2013 10 06 AM 150 Chapter 5 Picture This To add a second fitted line click the button in the Details field to bring up the Scatterplot Customize dialog Click for a new regres sion line this one using a log transformation on the Y variable Using the right specification sure fits the data a lot better Clever observation Did you notice that EViews cleverly solved for Y even though we specified log Y on the left of the series command EViews Illustrated book Page 150 Monday February 25 2013 10 06 AM Group Graphics 151 Scatter Plots and Distributions Scatter plots help us understand the joint distribution of two variables answering questions such as If variable one is high is variable two likely to be high as well We can add informa tion about the marginal distri butions of each variable by turning
336. rough two dimensional data But EViews tells us lots more than just slope and intercept In this chapter you ll see how easy it is to get parameter estimates plus a large variety of auxiliary statistics We begin our exploration of EViews regression tool with a quick look back at the NYSE vol ume data that we first saw in the opening chapter Then we ll talk about how to instruct EViews to estimate a regression and how to read the information about each estimated coef ficient from the EViews output In addition to regression coefficients EViews provides a great deal of summary information about each estimated equation We ll walk through these items as well We take a look at EViews features for testing hypotheses about regression coefficients and conclude with a quick look at some of EViews most important views of regression results Regression is a big subject This chapter focuses on EViews most important regression fea tures We postpone until later chapters various issues including forecasting Chapter 8 Forecasting serial correlation Chapter 13 Serial Correlation Friend or Foe and heteroskedasticity and nonlinear regression Chapter 14 A Taste of Advanced Estimation A First Regression Returning to our earlier examination of trend growth in the volume of stock trades we start with a scatter diagram of the logarithm of volume plotted against time EViews has drawn a straight line a re
337. rting changes the order in which the data is visually displayed The actual order in the workfile remains unchanged so analysis is not affected To restore the appearance to its original order sort using Observation Order and Ascending EViews Illustrated book Page 197 Monday February 25 2013 10 06 AM 198 Chapter 7 Look At Your Data As you can see Histo gram and Stats pro duces a histogram on the left and a panel of descriptive statistics on the right Let s start with the latter coming back to the picture part later The top of the statistics panel gives the sample in effect when the report was made and the num ber of observations If you compare this report with the one at the beginning of the chapter you might note that the smpl if command cut out two observa tions Comparing the maximum and minimum between the two reports we can deduce that one GPA of 39 and one GPA of 26 was eliminated Was it a good idea to eliminate these two observations This question can t be answered by statistical analysis you need to apply subject area knowledge In this case we might have chosen instead to correct the data by changing 39 to 3 9 and 26 to 2 6 Although one is left with the nagging question of whether there might really have been an applicant with a 0 26 GPA When we eliminated two grade observations by changing the sample we also cut out data for other series for these two individuals Their state of residenc
338. rustra tion I did the sensible thing and went and asked my wife s advice She told me Look at your data So I quickly pulled up a histo gram of the applicants grade point averages GPA Notice the one little data point all by its lonesome way off to the right According to the summary table the highest recorded GPA was 39 Since GPAs in American col leges are generally on a 4 0 scale it s a pretty good bet that a decimal point was omitted somewhere In this chapter we ll walk through a number of techniques for looking at your data Since the border between describing data and beginning an analy sis can be fuzzy some of the topics covered here are useful in data analysis as well Our dis cussion is split into univariate describing one variable at a time and multivariate describing several variables jointly Maybe it s easier to think of descriptive views of series and descriptive views of groups Hint Two important data descriptive techniques are covered elsewhere Graphing techniques are explored in Chapter 5 Picture This And while one of the very best techniques for looking at your data is to open a spreadsheet view and then look at it this doesn t require any instructions so past this reminder sentence we won t give any except for one little trick in the next section EViews Illustrated book Page 195 Monday February 25 2013 10 06 AM 196 Chapter 7 Look At Your Data Sorting Things Out
339. rvations in the Cps page are shown to the right Notice that the fourth and fifth person are both from Con necticut Even though the fourth person is a union member and the fifth person isn t they have the same value of AV_UNION We ve succeeded in attaching the state wide average unionization rate to each individual observa tion And the answer is Our regres sion results show that being a union member raises an individ ual s wages 23 5 percent Every additional percentage point of unionization in a state raises everyone s wages 2 65 percent EViews Illustrated book Page 265 Monday February 25 2013 10 06 AM 266 Chapter 9 Page After Page After Page Having Contractions Contracting data means mapping many data points into one Above when EViews contracted our data from individual to state level it used the default contraction method Mean EViews pro vides a variety of methods shown at the right in the Contraction method field Most of the contraction methods operate just as their names imply but No contractions allowed and Unique values are worth a bit of extra comment These last two options are primarily for error check ing Suppose as above that we want to link from state by state data to individual data There should only be one value from each state so the default contraction Mean just copies the state value However what if we had somehow messed up the state by state page so that there wer
340. rwritten ut EViews Illustrated book Page 316 Monday February 25 2013 10 06 AM Visual Checks 317 You ll note that the model does a good job of explaining volume after 1940 where the residuals fluctuate around zero and not such a good job before 1940 where the residuals look a lot like the dependent variable Visual Checks Every regression estimate comes with views to make looking at residuals easy If the errors are serially correlated then a large residual should generally be followed by another large residual a small residual is likely to be followed by another small residual and positive followed by positive and negative by negative Clicking the button brings up the Actual Fitted Residual menu Choosing Actual Fitted Residual Graph switches to the view shown to the right The actual dependent variable LOGVOL in this case together with the fitted value appear in the upper part of the graph and are linked to the scale on the right hand axis The residuals are plotted in the lower area linked to the axis on the left The residual plot includes a solid line at zero to make it easy to visu ally pick out runs of positive and Econometric hint We re treating serial correlation as a statistical issue Sometimes serial correlation is a hint of misspecification Although it s not something we ll inves tigate further that s probably the case here EViews Illustrated book Page 317 Monday Febr
341. ry 25 2013 10 06 AM 322 Chapter 13 Serial Correlation Friend or Foe dence in favor of serial correlation that we got from the Durbin Watson and the Breusch Godfrey More General Patterns of Serial Correlation The idea of first order serial correlation can be extended to allow for more than one lag The correlogram for first order serial correlation always follows geometric decay while higher order serial correlation can produce more complex patterns in the correlogram which also decay gradually In contrast moving average processes below produce a correlogram which falls abruptly to zero after a finite number of periods Higher Order Serial Correlation First order serial correlation is the simplest pattern by which errors in a regression equation may be correlated over time This pattern is also called an autoregression of order one or AR 1 because we can think of the equation for the error terms as being a regression on one lagged value of itself Analogously second order serial correlation or AR 2 is written More generally serial correlation of order p AR p is written When you specify the number of lags for the Breusch Godfrey test you re really specifying the order of the autoregression to be tested Moving Average Errors A different specification of the pattern of serial correlation in the error term is the moving average or MA error For example a moving average of order one or MA 1 would be writ
342. s 32 observations adding 54 55 order of 24 presentation of 58 small number of 219 one way tabulation 105 199 200 operations multiple 84 85 operators arithmetic 86 87 logical 86 87 order of 86 87 options EViews 404 409 ordinary least squares estimation 359 360 outliers 80 209 out of sample forecasting 227 overlapping graph lines 141 142 P pages accessing 239 240 creating 240 242 data conversion and 247 251 defined 26 deleting 243 earlier EViews versions and 243 importing into 241 mixed data 244 251 naming 241 242 243 saving 243 uses for 239 244 panels advantages 275 276 balanced unbalanced 278 defined 275 estimation 279 283 284 285 287 289 fixed effects 295 graphs 283 284 grouping 285 287 setting up 276 278 uses 271 272 versus pools 269 pasting data 38 52 53 55 56 250 251 See also importing pch function 108 perfect foresight models 380 period fixed effects 276 pie graphs 153 155 plus function 105 PNG 126 poff command 398 point forecasts 231 pon command 398 pools advantages disadvantages 294 characteristics 291 292 297 301 deleting series 303 estimation 294 304 308 example 291 293 294 exporting data 312 fixed effects 295 296 308 generating series 302 303 grouping 303 importing data 309 311 spreadsheet views 297 298 statistics 298 301 tools for 302 uses 271 272 versus panels 269 See also panels postscript files
343. s are given all is substituted for the first part We could pick out the last three weekdays with smpl if weekday gt 3 and weekday lt 5 We could pick out days in which the NASDAQ closed above 2 100 remember that the series Y is the NASDAQ closing price with smpl if y gt 2100 To select the days Monday and Wednesday through Friday in the first trading week of 2005 but only for those days where the NASDAQ closed above 2 100 type smpl 1 03 2005 1 03 2005 1 05 2005 1 07 2005 if y gt 2100 Hint It is a very common error to change the sample for a particular operation and then forget to restore it before proceeding to the next step At least the author seems to do this regularly EViews Illustrated book Page 97 Monday February 25 2013 10 06 AM 98 Chapter 4 Data The Transformational Experience Sample SMPLs Since the smpl command sets the sample you won t be surprised to hear that the sample command sets smpls The sample command creates a new object which stores a smpl In other words while the smpl command changes the active sample the sample command stores a sample specifica tion for future use Sample objects appear in the workfile marked with a icon Thus the command sample s1 1 03 2005 1 03 2005 1 05 2005 1 07 2005 if y gt 2100 stores in the workfile You can later reuse the sample specification with the com mand smpl s1 remembering that the specification is evaluated when used not when store
344. s first stacked on top of the data for the second country etc While Y and POP appear in the first row the country specific series names such as YCAN don t appear EViews Illustrated book Page 309 Monday February 25 2013 10 06 AM 310 Chapter 12 Everyone Into the Pool We want EViews to help out by attaching the identifiers in the first column to the series names beginning with Y and POP With the pool window active choose the menu Proc Import Pool data ASCII XLS WK Fill out the dialog with the names of the series to import as in the example to the right Hit and EViews will get everything properly attached Not surprisingly stacked by date is the flipped on the side version of stacked by cross section Here s an excerpt The same Proc Import Pool data ASCII XLS WK command works fine just choose the By Date radio button instead of By Cross section EViews Illustrated book Page 310 Monday February 25 2013 10 06 AM Getting Data In and Out of the Pool 311 Importing Stacked Data The Indirect Method Sometimes a little indirection makes life go more smoothly In the case at hand it s often easier to simply read your data into EViews by the methods you re already familiar with for example we might read in our data by drag and dropping the file onto the EViews desktop and clicking on to accept the defaults and then work with it in panel form see Chapter 11 Panel What s My Li
345. s for a nicer looking picture but can be misleading if it appears to report data that isn t there Hint You control whether lines are connected over NAs with the NA Handling option You can also use the broken sample options by including not isna x in the if part of your smpl statement EViews Illustrated book Page 180 Monday February 25 2013 10 06 AM Options Options Options 181 Options Options Options There are lots of options for fine tuning the appearance of your graphs The Graph Options dialog has seven sections each broken into pages filled with their own collection of details you can change Many of the options are obvious clicking a button marked lets you mess with the font right In this section we touch on the most important touch ups The Command Line Option Every option that can be set through dialogs can also be set by typing commands in the command pane In general it s a lot easier to use the dialogs The command line approach can be advantageous when you want to set the same options over and over If the tech niques covered in Templates for Success above and The Impact of Globalization on Intimate Graphic Activity below aren t powerful enough take a look at the Command and Program ming Reference Now back to our discussion of tweaking by dialog Graph Type From the Graph Type section you can change from one type of graph to another The only graph types that appear are tho
346. s indicate that the medians do not differ signifi cantly You can change the shading to a notch if you prefer as shown in the example below The short horizontal lines are called staples The upper staple is drawn through the highest data point that is no higher than the top of the box plus and analogously the lower staple is drawn through the lowest data point that is no lower than the bottom of the box minus The vertical lines connecting the staples to the box are called whis kers Data points outside the staples are called outliers Near outliers those no more than outside the staple are plotted with open circles and far outliers those further than outside the staple are plotted with filled circles There Can Be Less Than Meets the Eye Hint Boxplots tell you a lot about the data but don t jump to the conclusion that because a point is labeled an outlier that it s nec essarily got some kind of problem Even when data is drawn from a perfect normal distribution just over half a percent of the data will be identified as an outlier in a box plot 1 5 IQR 1 5 IQR 1 5 IQR 1 5 IQR EViews Illustrated book Page 209 Monday February 25 2013 10 06 AM 210 Chapter 7 Look At Your Data Boxplots By Categories Boxplots give quick visual com parisons of differ ent subpopulations In this plot we re again classifying GPA by Washing ton residence using the categorical graph tools We
347. s is just as happy to read a file from the web as it is to read a file from your disk Although you can t browse the web within EViews the way you can browse your disk you can enter a url i e a web address in the open file dialog Hint The counterpart to File Open Foreign Data as Workfile is File SaveAs A wide variety of file formats are accessible in the Save as type drop down menu part of which is shown to the right EViews Illustrated book Page 48 Monday February 25 2013 10 06 AM The Import Business 49 For example my friend Fred posts data on the 1 Year Treasury Constant Maturity Rate at http research stlouis fed org fred2 data WGS1YR txt In Inter net Explorer the data looks like the picture to the right Choose File Open For eign Data as Workfile and enter the url in the File name field Often the easiest way to grab a url is to copy it from the address bar of the web browser EViews Illustrated book Page 49 Monday February 25 2013 10 06 AM 50 Chapter 2 EViews Meet Data EViews does its usual nice job of interpreting the data Reading HTML The file above is a standard text file which happens to reside on the web More commonly files on the web are stored in HTML format HTML files can be stored on your local disk as well of course HTML files generally contain large amounts of formatting information which is invisible when displayed in a web browser EViews tries to
348. s tests 213 trend function 12 64 91 TRUE 86 87 tsls command 346 347 t statistics 67 69 76 213 2SLS two stage least squares estimation 345 347 two dimensional data analysis with panel See panels two stage least squares 2SLS estimation 345 347 U undo 19 84 243 unionization salary example 258 265 unit root problem 77 unit root testing 330 unlinking 254 unlocking windows 29 unobserved variables 275 276 unstacked spreadsheet view of pool 298 unstructured undated data 26 EViews Illustrated book Page 423 Monday February 25 2013 10 06 AM 424 Index untitled windows 29 up frequency conversion 247 249 upper function 104 usage view 111 V valmap command 109 value maps 109 112 variables adding 55 56 characteristics 24 dependent 14 63 65 70 dummy 99 endogenous 372 377 exogenous 372 377 explanatory 226 227 228 independent 14 63 65 66 left hand right hand side 65 limited dependent 349 351 model 372 377 naming 279 program 383 string 383 384 unobserved 275 276 variance decomposition 363 364 variance testing 211 vector autoregressions VARs 361 364 378 379 views graph 6 131 group 119 label 30 models 370 371 object 398 399 pools 297 298 series 4 6 11 13 30 spreadsheet 4 5 30 195 196 297 298 W Wald coefficient tests 75 78 360 warning messages customizing 404 web importing data from 48 51 weekday function 91 weighted least squares w
349. s that there aren t any omitted variables which are correlated with the included explanatory variables Omitted variables cause least squares estimates to be biased The usual problem is that if you don t observe a variable you don t have much choice but to omit it from the regression When 1 2 3 yit a bxit uit N T EViews Illustrated book Page 275 Monday February 25 2013 10 06 AM 276 Chapter 11 Panel What s My Line the unobserved variable varies across one dimension of the panel but not across the other we can use a trick called fixed effects to make up for the omitted variable As an example suppose y depends on both x and z and that z is unobserved but constant for a given coun try The regression equation can be written as where the variable z is stuffed inside the square brackets as a reminder that just like the error term u z is unobservable The trick of fixed effects is to think of there being a unique constant for each country If we call this constant and use the definition we can re write the equation with the unobservable z replaced by a separate intercept for each country EViews calls a cross section fixed effect The advantage of including the fixed effect is that by eliminating the unobservable from the equation we can now safely use least squares The presence of multiple observations for each country makes estimation of the fixed effect possible We could have j
350. s the contribution to the log likelihood function assuming that the errors are distributed independent Normal The last line announces to EViews that the contributions are in fact in LOGL1 The maximum likelihood coeffi cients are close to the coefficients estimated previously We ve gained formal estimates of the variances along with standard errors of the variance estimates The User s Guide devotes an entire chapter to the ins and outs of maxi mum likelihood estimation Addi tionally EViews ships with over a dozen files illustrating definitions of likelihood functions across a wide range of examples Hint We didn t really type in that long seasonal component We copied it from the representations view of the earlier least squares results pasted and did a little judi cious editing EViews Illustrated book Page 356 Monday February 25 2013 10 06 AM System Estimation 357 System Estimation So far all of our estimation has been of the one equation at a time variety System estima tion in contrast estimates jointly the parameters of two or more equations System estima tion offers three econometric advantages at the cost of one disadvantage The first plus is that a parameter can appear in more than one equation The second plus is that you can take advantage of correlation between error terms in different equations The third advan tage is that cross equation hypotheses are easily tested The disadvantage is t
351. s the obvious set of choices for display ing or not displaying tick marks on the axes Honest graph alert In Chapter 5 page 157 we saw a bar graph comparing wages of union and non union workers Automatic selection chose a pretty but substantively question able endpoint for the y axis Here s a better ver sion where we ve User specified a lower limit of zero EViews Illustrated book Page 185 Monday February 25 2013 10 06 AM 186 Chapter 6 Intimacy With Graphic Objects Top and Bottom Axes The choices for marking the top and bottom axes vary depending on whether the horizontal scale displays num bers in which case the choices are essentially the same as the ones we ve just seen or if as is more common the bottom scale shows dates In the latter situa tion the choices are the ones appropriate to dates and the exact choices depend on the frequency of the workfile Options for the date scale can be found on the Obs Date axis page The Date format dropdown provides a variety of fairly self explanatory choices including Custom for when you want to roll your own See the User s Guide Observations to label similarly provides both a selection of built in and custom options You can cause grid lines to be displayed from the Grid Lines page The Obs Date axis grid lines field lets you customize the interval of grid lines for the date axis Hint Just as numerical scales sometimes appea
352. s to be enclosed in quotes We ve already seen one such example The function mean accepts a sample specification as an optional second argument Following along from the preceding example we could compute smpl all EViews Illustrated book Page 99 Monday February 25 2013 10 06 AM 100 Chapter 4 Data The Transformational Experience series overallmean mean y series samplemean1 mean y s2 series samplemean2 mean y first 1 last if y y 1 OVERALLMEAN gives the mean of Y taken over all observations SAMPLEMEAN1 takes the mean of those observations included in S2 as does SAMPLEMEAN2 Data Types Plain and Fancy Series hold either numbers or text That s it Except that sometimes numbers aren t numbers they re dates And sometimes the numbers or text you see aren t there at all because you re looking at a value map instead We ll start simply with numbers and text and then let the discussion get a teeny bit more complicated Numbers and Letters As you know the command series creates a new data series each observation holding one number Series icons are displayed with the series icon EViews stores numbers with about 16 digits of accuracy using what computer types call double precision Computer hint Computer arithmetic is not perfect On very rare occasion this mat ters Data hint Data measurement is not perfect On occasion this matters a lot EViews Illustrated book Page 100
353. saved and that Untitled ones aren t This design lets you try out things without cluttering the workfile Related hint You can use the EViews menu item Options General Options Win dows Window Behavior to control whether you get a warning before closing an Unti tled window See Chapter 18 Optional Ending Hint for the terminally obedient For goodness sakes don t really enter all the data at the beginning of the chapter You ll be bored out of your mind Just type in a few num bers until you re comfortable moving around in the window EViews Illustrated book Page 29 Monday February 25 2013 10 06 AM 30 Chapter 2 EViews Meet Data Label View We know that EViews provides several different views for looking at a series We enter data in the spreadsheet view and if we need to make a change we can come back to the spreadsheet view to edit existing data Use the button to reach the label view where space is pro vided for you to enter a description source etc EViews automatically fills in the name and date the series was last updated The other fields are optional EViews uses the Display Name for labeling output so it s well worth filling out this field Make the label long enough to be meaningful but short enough to fit in scarce space on a graph legend EViews will occasionally make an entry in the Remarks field When you start making transformations to a series a History field is added with notes on the
354. se that are permissible For example if you re looking at a single series you won t be offered a scatter plot Hint The Basic type page for a frozen graph that has updating off offers a limited set of options typically far fewer than are available for a graphical view of a series or group EViews Illustrated book Page 181 Monday February 25 2013 10 06 AM 182 Chapter 6 Intimacy With Graphic Objects Frame amp Size The Frame amp Size section is the place for set ting options that are essentially unrelated to the data being graphed The Color amp Border page lets you set specifications for the frame itself On the left you can set colors for the area inside the frame Frame fill and the area outside the frame Background On the right you can set aspects of the frame border even elimi nating the border entirely if you wish The Size amp Indents page of the Frame amp Size section lets you set a margin for the graph inside the frame EViews Illustrated book Page 182 Monday February 25 2013 10 06 AM Options Options Options 183 The left side of the Size amp Indents page lets you set the Frame size except that the frame size on the screen doesn t change What this field really lets you choose is the shape of the frame sometimes called the aspect ratio within the size of the existing win dow So if you choose 2 inches high and 8 inches wide you get a really wide fr
355. sed sam ple size has raised the t statistic on population growth from 3 to 5 EViews Illustrated book Page 294 Monday February 25 2013 10 06 AM Getting Your Feet Wet 295 There s nothing special about moving the constant term into the Cross section specific coefficients field You can do the same for any variable you think appropriate Fixed Effects Okay that first sentence was a fib There is something spe cial about the constant The cross section specific con stant picks up all the things that make one country differ ent from another that aren t included in our model Such differences occur so fre quently that EViews has a built in facility for allowing for such country specific con stants Country specific con stants are called fixed effects Push again take the constant term out of the spec ification entirely and set Esti mation method to Cross section Fixed Observation about life as a statistician Running estimates until you get results that accord with prior beliefs is not exactly sound practice The risk isn t that the other guy is going to do this intentionally to fool you The risk is that it s awfully easy to fool yourself unintentionally Hint The econometric issues surrounding fixed effects in pools are the same as for panels See Chapter 11 Panel What s My Line EViews Illustrated book Page 295 Monday February 25 2013 10 06 AM 296 Chapter 12 Everyone Into the Pool
356. series contain ing on education and a dummy variable for each gender produces nicely labeled output More on the expand function under Expand the Dummies In fact you can use the value labels in editing a series Be sure the Display type menu is set to Default EViews checks to see if an entry matches any of the labels in the value map If so the corre sponding value is entered If not the new entry is used directly As a consequence you can use either the label or underlying code to enter a new value so long as the underlying code hasn t been used to label some other value Since a value map is just a translation from underlying code to appearance you re free to use the same value map for multiple series To help keep track of which series is using a given value map the Usage view of a value map shows a list of all the series currently attached Hint Labels in the editor are case sensitive e g male translates to 0 but Male translates to NA Hint EViews doesn t care what you use for a label It will cheerfully let you label the number 0 with the value map 1 Don t For further discussion see Origin of the Species Hint If you load data created in another stat program that has its own version of value maps EViews will create value maps and correctly hook them up to the relevant series However there s no general method using values in an alpha series as labels in a value map
357. sion on French data reported above except that this time we ve combined the data for all six countries Let s see what s changed First we have 280 obser vations instead of 50 Second the reported effect of population has switched sign The French only result was negative as theory pre dicts The pooled result is positive EViews Illustrated book Page 293 Monday February 25 2013 10 06 AM 294 Chapter 12 Everyone Into the Pool Everyone Into the Pool May Not Be Fun The advantage of pooling data is that a great deal of data is brought to bear on the problem The potential disad vantage is that a simple pool forces the coefficients to be identical across countries Does this make sense in our example We probably do want the coefficient on popu lation growth to be the same for each country because the theory isn t of much use if population growth doesn t have a predictable effect In contrast there s no reason for the intercept to be the same for each country We know that countries have different levels of GDP for reasons unrelated to population growth Let s retry this estimate with an individual intercept for each country Go back to the estimation dialog and move the constant from the Common coefficients field into Cross section specific coefficients Now we re asking for a separate intercept for each country The estimated effect of population growth is negative as we had expected And the increa
358. statistic So this hypothesis is also easily rejected Hint The p value reported by EViews is computed for a two tailed test If you re inter ested in a one tailed test you ll have to look up the critical value for yourself EViews Illustrated book Page 76 Monday February 25 2013 10 06 AM Hypothesis Testing 77 EViews is happy to test a hypothesis involving multiple coefficients and nonlinear restrictions To test that the sum of the first two coefficients equals the product of the sines of the second two coefficients and to emphasize that EViews is per fectly happy to test a hypothesis that is complete nonsense enter c 1 c 2 sin c 3 sin c 4 Not only is the hypothesis nonsense appar ently it s not true Econometric theory warning If you ve studied the advanced topic in econometric the ory called the unit root problem you know that standard theory doesn t apply in this test although the issue is harmless for this particular set of data Take this as a reminder that you and EViews are a team but you re the brains of the outfit EViews will obediently do as it s told It s up to you to choose the proper procedure EViews Illustrated book Page 77 Monday February 25 2013 10 06 AM 78 Chapter 3 Getting the Most from Least Squares A good example of a hypothesis involving multiple restrictions is the hypothesis that there is no time trend so the coefficients on both and equal zero
359. stimated standard deviation of the error term is 0 97 Five other elements Sum squared residuals Log likelihood Akaike info criterion Schwarz criterion and Hannan Quinn criter are used for making statistical comparisons between two different regressions This means that they don t really help us learn anything about the regression we re working on rather these statistics are useful for deciding if one model is better than another For the record the sum of squared residuals is used in computing F tests the log likelihood is used for computing likelihood ratio tests and the Akaike and Schwarz criteria are used in Bayesian model comparison The next two numbers Mean dependent var and S D dependent var report the sample mean and standard deviation of the left hand side variable These are the same numbers you d get by asking for descriptive statistics on the left hand side variables so long as you were using the sample used in the regression Remember EViews will drop observations from the estimation sample if any of the left hand side or right hand side variables are NA i e missing The standard deviation of the dependent variable is much larger than the standard error of the regression so our regression has explained most of the variance in log volume which is exactly the story we got from looking at the R squared Why use valuable screen space on numbers you could get elsewhere Pri
360. strated book Page 406 Monday February 25 2013 10 06 AM Window Behavior 407 Window Behavior When EViews opens a new window the win dow title is Untitled unless you ve explicitly given a name Named objects are stored in the workfile untitled objects aren t Com mands like ls and show create untitled objects So does freezing an object If you do a lot of exploring you ll create many such untitled objects Unchecking Warn on close lets you close throw aways without having to deal with a delete confir mation dialog On the other hand once in a while you ll delete something you meant to name and save As suggested earlier you ll probably want to uncheck most of the options in Allow only one untitled Doing so lets you type a sequence of ls commands for example without having to close windows between commands The downside is that the screen can get awfully cluttered with accumulated untitled windows Keyboard Focus The radio button Keyboard focus directs whether typed characters are sent by default to the command pane or to the currently active window Most people leave this one alone but if you find that you re persistently typing in the command pane when you meant to be edit ing another window try switching this button and see if the results are more in line with what your fingers intended EViews Illustrated book Page 407 Monday February 25 2013 10 06 AM 408 Chapter 18 Optiona
361. t Copy For example to move Agriculture and Engineering to the bottom select the two rows as shown and select Copy Then making certain that edit mode for the group is on paste into rows at the bot tom of the workfile Question Can I insert observations in the middle of my file instead of at the end Response Nope Further Response Yep EViews Illustrated book Page 54 Monday February 25 2013 10 06 AM Adding Data To An Existing Workfile Or Being Rectangular Doesn t Mean Being Inflexible 55 Finally clear out the area you copied from by selecting each cell in turn and hitting the Delete key Adding new variables Adding a new variable or variables is relatively easy Think of adding a blank column to the right of the data rectangle If the new variable is in an existing workfile or if you can arrange to get it into one add ing the variable into the destination workfile is a cinch EViews treats each series as a uni fied object containing data frequency sample label etc Open the source workfile select the series you want and select Copy Open the destination workfile and Paste All done Suppose for example you have data on the clipboard that you want to add to an existing workfile Use the Paste as new Workfile procedure we talked about earlier in the chapter to create a new workfile Then Copy Paste series from the new workfile into the desired desti nation workfile If the series you want to add
362. t can Hint In case you were wondering there s no way to automatically label individual slices EViews Illustrated book Page 154 Monday February 25 2013 10 06 AM Let s Look At This From Another Angle 155 If you want to use color but still get acceptable monochrome ren derings change the colors in the Fill Areas section in the Graph Elements group of the Graph Options dialog to ones with dif ferent darkness levels Here we ve used blue pink and white You can also select differ ent Grey shade levels which operate independently of the color choice when EViews ren ders in black and white Let s Look At This From Another Angle To twist a graph on it s side choose Rotated obs axis on left in the Orientation combo of the Graph Type dialog Below is a rotated ver sion of the bar graph we saw on page 139 Hint The same issue arises in any graph with adjacent filled areas You can use the same trick in any graph EViews Illustrated book Page 155 Monday February 25 2013 10 06 AM 156 Chapter 5 Picture This To Summarize To visually summarize your data change the Graph data dropdown in the Details field to a summary statistic of your choice For example here s a bar graph showing the median level of wages and salaries for U S workers in 2004 cpsmar2004extract wf1 Pretty boring eh Even if you re fascinated by wage distributions that s a pretty boring graph All the choic
363. t chapter we ll go through all the required steps to set up a workfile from scratch Now that a workfile s loaded EViews looks like this Our workfile contains information about quarterly average daily trading volume on the New York Stock Exchange NYSE There s quite a bit of information with over 400 observations taken across more than a century The icon indicates a data series stored in the workfile Viewing an individual series If you double click on the series VOLUME in the workfile window you ll get a first peek at the data Right now we re looking at the spreadsheet view of the series VOLUME The spread sheet view shows the numbers stored in the series On average in the first quarter of 1888 159 006 shares were traded on the NYSE The numbers for VOLUME were recorded in millions Interesting but per haps a little outdated for understanding today s market EViews Illustrated book Page 4 Monday February 25 2013 10 06 AM Viewing an individual series 5 Scroll down to the bottom of the window to see the latest datum in the series 1 67 bil lion shares were traded on an average day in the first quarter of 2004 Quite a change The more ways we can view our data the better EViews provides a collection of Views for each type of object that can appear in a workfile Object is a com puter science buzz word meaning thin gie The figures above are examples of the spreadsheet vi
364. t editor before bringing it into EViews Dentistry 44214 Medicine 43160 Law 40670 Agriculture 36879 Engineering 35694 EViews Illustrated book Page 51 Monday February 25 2013 10 06 AM 52 Chapter 2 EViews Meet Data Double click on Range in the upper pane of the workfile win dow or choose Proc Struc ture Resize Current Page to bring up the workfile structure dialog This dialog lets you change the range and or the structure of the workfile Be care ful not to change the structure by accident In our example we want an unstructured workfile with 6 observations so the filled out dialog looks like this Open a group window containing DISCIPLINE and SALARY You can see that a row with no data has been added at the bottom SALARY is marked NA for not available and DISCIPLINE is an empty string An empty string just looks like a blank entry in the table Now type in the new observation It s okay for some of the new entries to be left as NA just as there can be NAs in the existing data Copy Paste You could of course have added a thousand new observations just as easily as one Typing 1 000 observations would be rather tedious though In contrast Copy Paste isn t any harder for 1 000 observations than it is for one Go to the computer file holding your data Copy the data you wish to add being sure that you ve selected a rectangle of data In EViews open a group with the desired varia
365. t is the coefficient number C 1 is the first element of C C 2 is the second element of C and so on EViews Illustrated book Page 342 Monday February 25 2013 10 06 AM Nonlinear Least Squares 343 The results haven t changed but the names given to the estimated coeffi cients have As a side effect the results are stored in ALPHA and BETA not in C Notice that the slope coefficient has been stored in BETA 1 and that since we didn t reference BETA 2 nothing has been stored in it Coefficient vectors aren t just for storing results They can also be used in computations For example to compute squared residuals one could type series squared_residuals volume alpha 1 beta 1 trend 2 Hint Since every new estimate specified with a series list replaces the values in C it makes sense to use a different coefficient vector for values you d like to keep around Hint It would of course been easier in this example to enter series squared_residuals resid 2 But then we wouldn t have had the opportunity to demonstrate how to use coefficients in a computation EViews Illustrated book Page 343 Monday February 25 2013 10 06 AM 344 Chapter 14 A Taste of Advanced Estimation Making It Really Nonlinear To estimate a nonlinear regression enter a nonlinear formula in the ls command For example to estimate the equation use the command ls volume c 1 1 trend c 2 Notice the big improvement in
366. t the cost of drop ping the information that the rate quotes come from the sec ondary market In general detailed information is probably better in a footnote or figure cap tion But the choice of course is yours Graph Elements The Graph Elements section contains options for specific graph types Lines amp Symbols The Attributes field on the right in the Lines amp Symbols page is the place to pick colors and pat terns for the lines and symbols for each series Click on the num bered lines at the far right to select the series to adjust Note that the legend label 3 MONTH TREASURY appears at the bottom of the Attributes field to identify the selected series The provided drop down menu lets you choose whether to use a line a symbol or both for each series The right most field displays the representation for each series showing how the series will be rendered in color and how it will be rendered in black and white The default pattern uses the colors shown in the Color column of the Attributes field for color rendering and EViews Illustrated book Page 188 Monday February 25 2013 10 06 AM Options Options Options 189 the line patterns shown under B amp W for black and white rendering The Pattern use radio button on the left side of the page specifies whether to use the default Auto choice or force all rendering to solid or all to pattern See the discussion under A Little Light Custom i
367. t what they say Justification lets you specify left right or center justification when there are multiple lines in a single label The button opens the Font Color dialog which has pretty standard options The checkbox Text in Box and the Box fill color and Frame color dropdowns are also used in the obvious way The Position field is worth a comment or two Left Rotated and Right Rotated tilt text 90 degrees so The default position User sticks the text inside the frame of the graph itself The location is measured in virtual inches see Frame amp Size on page 182 with positive numbers moving down and to the right Frankly sometimes it s easiest to just stick the text any old place and then use the mouse to re position it EViews Illustrated book Page 169 Monday February 25 2013 10 06 AM 170 Chapter 6 Intimacy With Graphic Objects The figure to the right shows labels in a variety of positions Take note that the default position ing is User You ll probably find yourself most often adding text to put a title on the top or bottom of the graph Since EViews fre quently sticks a legend at the bot tom of the graph you may find it best to place the title at the top EViews goes to a lot of trouble to make text look nice Each letter is carefully placed for the best appearance A side effect of this extra care is that other programs may have trouble editing text included inside graphs copied and paste
368. ta are still annual from 1950 through 2000 but the workfile is structured to contain two cross sections Canada and the U S All the population measurements are in the single series POP Here s a quick peek at the data For the pooled data on the left we see the first few observations for population for Canada and the U S each in its own series The two series have intentionally and usefully similar names but nothing is fundamentally different from what we ve seen before EViews Illustrated book Page 270 Monday February 25 2013 10 06 AM Nuances 271 The panel on the right shows the single series POP But look at the row labels they show both the country name and the year The rows shown we ve scrolled to roughly the mid dle of the series are the Canadian data for the end of the sample followed by the U S data for the early years In a panel the data for different countries are combined in a single series We get the all the observations for the first country first followed by all the observa tions for the second country Unlike pools the panels do introduce a fundamentally new data structure You can think of a pool as a sort of ber group A pool isn t a group of series but it is a set of identifiers that can be used to bring any set of series together for processing If our work file had also included the series GDPCAN and GDPUSA the same ISOCODE pool that con nects POPCAN and POPUSA would also connect GDP
369. ta yet but let s dissect what EViews starts you off with The initial workfile window looks something like the picture to the right The title bar shows the name of the workfile Since we didn t enter a name for the workfile in the dialog EViews uses UNTITLED in the title bar The workfile window has buttons at the top and tabs at the bottom The buttons provide menus linked to each EViews window type The tabs mark pages essentially workfiles within a workfile We ll come back to pages in Chapter 9 Page After Page After Page they re particularly useful for holding sets of data with different indices Hint Alternatively we can enter wfcreate u in the command pane Hint If you like the workfile can be created with the single command wfcreate u 28 EViews Illustrated book Page 26 Monday February 25 2013 10 06 AM Time to Type 27 Let s look at the main window area which is divided into a upper pane holding information about the workfile and a lower pane displaying information about the objects series equa tions etc that are held in the workfile Range tells you the identifying numbers or dates of the first and last observation in the workfile 1 and 28 in this example as well as the count of the number of observations Sample describes the subset of the observations range being used for current operations Since all we ve done so far is to set up a workfile with 28 observations both Ra
370. te a space to separate EViews Illustrated book Page 6 Monday February 25 2013 10 06 AM Looking at different samples 7 the dates and the special code last to pick up the last date available in the workfile When you ve changed the Sample dialog as shown in the second figure hit The Line Graph view changes to reflect the new sample Note how the date scaling on the horizontal axis has changed Previously we could fit only one label for each decade This close up view gives a label every quarter For example III with 2002 below it means year 2002 third quarter which is to say July September 2002 Because the sample is so much more homoge nous we can now see lots of short run up and down spikes We could ve also changed the sample by resizing and dragging the slider bar at the bottom of the window Notice that in the first graph the slider bar was very wide extending the width of the entire sam ple Now its size and position reflect the new sample which shows a small portion of the latter part of our data Remember when we changed the sample on the view we have not changed the underlying data just the portion of the data that we re viewing at the moment The complete set of data remains intact ready for us to use any time we d like Hint Ranges of dates in EViews are specified in pairs so 2001q1 last means all dates starting with the first quarter of 2001 and ending with the last
371. te the text display See Chapter 5 Picture This You can also Edit Copy the text in the statistics panel and then paste the text into your word processor Hint There are relatively few places in econometrics where normality of the data is important In particular there is no requirement that the variables in a regression be normally distributed I don t know where this myth comes from EViews Illustrated book Page 199 Monday February 25 2013 10 06 AM 200 Chapter 7 Look At Your Data Tabulation of GPA provides lots of information It also illustrates a com mon problem too many categories This is why the Tabulate Series dialog defaults provides binning control Binning Control The Group into bins if field is a three part control over grouping individual values into bins Checking of values tells EViews to create bins if there are more than the specified number of val ues and checking Avg count means to create bins if the average count in a category is less than specified Max of bins not surprisingly sets the maxi mum number of bins Sometimes you need to play around with these options to get the tabulation that best fits your needs As an example here s a GPA tabula tion that shows broad categories It s now easy to see that 15 percent of applicants had below a 3 0 average and 10 appli cants 0 61 percent of the applicant pool did report GPAs above 4 0 EViews Illustrated book Page 200
372. tead of a new separate workfile We ll get to the other two options later in the chapter EViews Illustrated book Page 240 Monday February 25 2013 10 06 AM Creating New Pages 241 Creating A New Page From Scratch Choosing Specify by Frequency Range brings up the familiar Workfile Create dia log You ll see that the field for the workfile name is greyed out since you re creating a page rather than a workfile Creating A New Page From Existing Data Paste from Clipboard as Page creates an untitled page by reading the data on the clipboard Paste from Clipboard as Page is analogous to right clicking in an empty area of the EViews window and choosing Paste as new Workfile except that you get a page within the existing workfile rather than a new separate workfile You can achieve the same effect without using the clipboard by dragging a source file and dropping it on the New Page tab A plus sign will appear when your cursor is over an appropriate area Load Workfile Page reads an existing workfile from the disk and copies each page in the disk workfile into a separate page in the active workfile Creating A New Page Based On An Existing Page Copy Extract from Current Page brings up a cascading menu with the two choices shown to the right The two options do essentially the same thing We ll defer a discus Hint The Workfile Create dialog always gives you the opportunity to name the page being created
373. ted as a pool while large anonymous e g survey respondent 17529 cross sections may be better ana lyzed as a panel EViews Illustrated book Page 271 Monday February 25 2013 10 06 AM 272 Chapter 10 Prelude to Panel and Pool If the number of cross sections is really large you pretty much have to use a panel What s really large Remember that in a pool each cross section element has a series for every variable If the cross section is large enough that typing the names of all the countries people etc is painful you should probably use a panel So What Are the Benefits of Using Pools and Panels We ll spend the next two chapters answering this question The big answer is that you can control for common elements across observations or not as you choose The smaller answer is that all sorts of data manipulation are made easier because EViews understands how different observations are tied together As a quick example from the poolside here s a set of descriptive statistics done for each country for the whole time series What s more it would have been no more trouble to produce these statistics for 20 countries than it was for two With one click of a differ ent button we can get descriptive statistics done for each year for all two countries grouped together Quick P review If you have a cross section of time series put them into a pool If you have repeated obser vations on cross sectio
374. ters of our sample based on the model including an AR 1 error Since we know what actually happened in that period we can compare our forecast with reality to see how well we ve done The first quarter of our forecast period is 2002q2 First EViews will multiply the right hand side variables for 2002q2 by their respective estimated coefficients Then EViews adds in the contribution of the AR 1 term times the residual from 2002q1 the final period before the forecast began The second period of our forecast is 2002q3 Now EViews will multiply the right hand side variables for 2002q3 by their respective estimated coefficients Then there s a choice should we add in times the residual from 2002q2 or should we use times the residual from 2002q1 The former is called a static forecast and the latter a dynamic forecast The static forecast uses all information in our data set while the dynamic forecast uses only information through the start of the estimation sample The static forecast uses the best available information so it s likely to be more accurate On the other hand if we were truly forecasting into an unknown future dynamic forecasting would be the only option Static forecasting requires calculation of residuals during the forecast period If you don t know the true values of the left hand side variable you can t do that Therefore dynamic forecasting is generally a better test of how well multi period forecasts would work
375. that our analysis had covered a dozen or so variables and that the contract required cross tabulations of all the variables not just those that the research team thought mattered And that we d better comply The letter was turned over to me I found my mentor the next day and pointed out that the sponsoring agency was asking for about one million pages of printout He laughed Told me to mail them the first thousand pages without comment and said we d never hear back I did and we didn t Hint Sometimes you can get a more useful table by printing in landscape rather than portrait mode Hint Sometimes when there are too many tables to manage visually you should stop and think about whether there are also too many tables to help you learn anything EViews Illustrated book Page 221 Monday February 25 2013 10 06 AM 222 Chapter 7 Look At Your Data EViews Illustrated book Page 222 Monday February 25 2013 10 06 AM Chapter 8 Forecasting Prediction is very difficult especially about the future Niels Bohr Think what an easier time Bohr would have had if he d had EViews instead of just a Nobel prize in physics Truth be told the design of a good model on which to base a forecast can be very diffi cult indeed EViews role is to handle the mechanics of producing a forecast it s up to the researcher to choose the model on which the forecasts are based We ll start off with an example of just how remark
376. the second quarter for 1933 The actual value looks out of line with the surrounding val ues Perhaps this was a really unusual quarter on the NYSE or maybe someone even wrote down the wrong numbers when putting the data together Grabbing the Residuals Since there is one residual for each observation you might want to put the residuals in a series for later analysis Fine All done Without you doing anything EViews stuffs the residuals into the special series after each estimation You can use RESID just like any other series EViews Illustrated book Page 80 Monday February 25 2013 10 06 AM Quick Review 81 Quick Review To estimate a multiple regression use the ls command followed first by the dependent vari able and then by a list of independent variables An equation window opens with estimated coefficients information about the uncertainty attached to each estimate and a set of sum mary statistics for the regression as a whole Various other views make it easy to work with the residuals and to test hypotheses about the estimated coefficients In later chapters we turn to more advanced uses of least squares Nonlinear estimation is covered as are methods of dealing with serial correlation And predictably we ll spend some time talking about forecasting Resid Hint 1 That was a very slight fib EViews won t let you include RESID as a series in an estimation command because the act of estimation changes the
377. ther EViews Illustrated book Page 165 Monday February 25 2013 10 06 AM 166 Chapter 5 Picture This You ll note that this last picture has been messed with some colors titles and axes have changed Such messing around techniques are covered in Chapter 6 Intimacy With Graphic Objects Quick Review and Look Ahead EViews makes visually pleasing graphs quite easily All you need do is open a series or a group and choose from the wide variety of graph types available A variety of summary sta tistics can be graphed as easily as raw data You can also have EViews make individual plots for data falling into different categories Customizing graphs by adding text or chang ing colors is similarly easy If your visual needs have been satisfied then this chapter is all you need to know about graphics If more control is your style then the next chapter will make you very happy EViews Illustrated book Page 166 Monday February 25 2013 10 06 AM Chapter 6 Intimacy With Graphic Objects EViews does a masterful job of creating aesthetically pleasing graphs But sometimes you want to tweak the picture to get it just so choosing custom labeling axes colors etc In this chapter we get close up and intimate with EViews graphics We organize our tweaking exploration around four buttons in the graph window and ordered from easiest to most sophisticated This last button brings up the Graph Options dialog
378. they The pattern suggests that an AR 1 isn t a good enough specification which we already suspected from other evidence Here s the analogous correlogram from the ARMA 2 1 model we esti mated earlier In this more general model the theoretical correlogram and the empirical correlogram are much closer The richer specifica tion is probably warranted The Impulse Response Function Including ARMA errors in forecasts sometimes makes big improvements in forecast accu racy a few periods out The further out you forecast the less ARMA errors contribute to forecast accuracy For example in an AR 1 model if the autoregressive coefficient is esti mated as 0 9 and the last residual is then including the ARMA error in the forecast adds Nomenclature Hint The theoretical correlogram corresponding to the estimated ARMA parameters is sometimes called the Autocorrelation Function or ACF eT EViews Illustrated book Page 326 Monday February 25 2013 10 06 AM Forecasting 327 in the first three forecasting periods As you can see the ARMA effect gradually declines to zero You can see that two elements determine the contribution of ARMA errors to the forecast the value of the last residual and how quickly the weights decline The value of the last residual depends on the starting date for the forecast but the weights can be plotted using ARMA Structure Impulse Response The weights are multiplied either by the standard
379. though the youknowwhatimeant function is eagerly awaited in a future release In the meantime the One Way Tabulation view see Chapter 7 Look At Your Data gives a list of all the unique values of a series This offers a quick check for various spellings for alpha series where a limited number of val ues are expected as is true of state names Uncheck the boxes in the Tabulate Series dialog except for Show Count The view that pops up gives a nicely alphabetized list of spellings One more string function is very useful but doesn t look like a function When used between alpha series means concatenate Thus the command a b c gives the string abc Hint Typing an equal sign fol lowed by an expression turns EViews into a nice desk calcu lator Results are shown in the status line at the bottom of the EViews window EViews Illustrated book Page 105 Monday February 25 2013 10 06 AM 106 Chapter 4 Data The Transformational Experience Can We Have A Date Technically EViews doesn t have a date type Instead EViews has a bunch of tools for interpreting numbers as dates If all you do with dates is look at them there s no need to understand what s underneath the hood This section is for users who want to be able to manipulate date data The key to understanding EViews representation of dates is to take things one day at a day An observation in a date series
380. ticity Tests view of an equation offers variance heteroskedasticity tests includ ing two variants of the White heteroske dasticity test The White test is essentially a test of whether values of the right hand side variables and or their cross terms etc help explain the squared residuals To perform a White test with only the squared terms no cross terms you should uncheck the Include White cross terms box Here are the results of the White test without cross terms on our currency growth equation The F and statistics reported in the top panel decisively reject the null hypothesis of homoskedasticity The bottom panel only part of which is shown shows the auxil iary regression used to compute the test statistics x1 2 x1 x2 x2 2 x2 EViews Illustrated book Page 339 Monday February 25 2013 10 06 AM 340 Chapter 14 A Taste of Advanced Estimation Heteroskedasticity Robust Standard Errors One approach to dealing with het eroskedasticity is to weight observa tions such that the weighted data are homoskedastic That s essen tially what we did in the previous section A different approach is to stick with least squares estimation but to correct standard errors to account for heteroskedasticity Click the button in the equation window and switch to the Options tab Select either White or HAC Newey West in the dropdown in the Coefficient covariance matrix group As an example we
381. trated book Page 175 Monday February 25 2013 10 06 AM 176 Chapter 6 Intimacy With Graphic Objects Bold wide and English labels in templates You will also be given the choice of applying the Bold Wide or English labels modifiers Bold modifies the template settings so that lines and symbols are bolder thicker and larger and adjusts other characteristics of the graph such as the frame to match Wide changes the aspect ratio of the graph so that the horizontal to vertical ratio is increased and English labels modifier changes the settings for auto labeling the date axis so that labels that use month formatting will default to English month names Jan Feb Mar instead of M1 M2 M3 Predefined Templates EViews comes with a set of predefined templates that provide attractive looking graphic styles These appear on the left in the Template selection field You can add any graph you like to the predefined list so that it will be globally avail able by going to the Manage templates page of the Tem plates amp Objects section Then EViews Illustrated book Page 176 Monday February 25 2013 10 06 AM Point Me The Way 177 use the dialog to add the graph to the Predefined templates list Note that predefined tem plates don t include text or lines shades from the original graph You can make any template the default for all new graphs by going to the Options Graphics Defaults men
382. tributed uniformly between A and B This means that rnd is a synonym for runif 0 1 We can make up an example with series x runif 0 2 show x dunif x 0 2 cunif x 0 2 qunif cunif x 0 2 0 2 which randomly resulted in the data shown to the right The first col umn X is a random number ran domly distributed between 0 and 2 The second column gives the pdf for X which for this distribution always equals 0 5 The third column gives the cdf Just for fun the last column reports the inverse cdf of the cdf which is the original X just as it should be A variety of probability distributions are discussed in the Command and Programming Refer ence Probably the most commonly used are norm for standard normal and tdist for Student s t Here are a few examples qtdist 95 30 1 69726 qtdist 0 05 2 30 2 04227 qnorm 025 1 95996 Quick Review Data in EViews can be either numbers or text A wide set of data manipulation functions are available In particular the expected set of algebraic manipulations all work as expected You can use numbers to conveniently represent dates EViews also provides control over the EViews Illustrated book Page 114 Monday February 25 2013 10 06 AM Quick Review 115 visual display of data Value maps and display formats play a big role in data display In par ticular value maps let you see meaningful labels in place of arbitrary numerical cod
383. trols 386 A Rolling Example 387 Quick Review 388 Appendix Sample Programs 388 CHAPTER 17 ODDS AND ENDS 393 How Much Data Can EViews Handle 393 How Long Does It Take To Compute An Estimate 393 Freeze 393 A Comment On Tables 395 Saving Tables and Almost Tables 396 Saving Graphs and Almost Graphs 396 Unsubtle Redirection
384. tting Up Panel Data 277 The figure to the right shows observations 1597 through 1610 which happen to be the last few observations for the Central African Republic and the first few observations for Canada To us humans it s clear that these are observations for and for and then starting over with In order to set up a panel structure we need to share this kind of understanding with EViews Structuring a panel workfile To change from a regular to a panel structure use the Workfile structure dialog Double click on Range in the upper pane of the workfile window or use the menu Proc Structure Resize Cur rent Page Choose Dated Panel for the Workfile struc ture type and then specify the series containing the cross section i and date t identi fiers EViews re organizes the workfile to have a panel structure the re organized workfile is available in the EViews web site as PWT61PanelExtract wf1 EViews announces the panel aspect of the workfile structure by changing the Range field in the top panel of the workfile window We now have 208 cross sections for data from 1950 through 2000 i CAF CAN t 1994 2000 t 1950 EViews Illustrated book Page 277 Monday February 25 2013 10 06 AM 278 Chapter 11 Panel What s My Line More information about the structure of the workfile is available by push ing the button and choosing Sta tistics from the menu Let s take another l
385. tton to switch the model to the variables view We see that CONS and Y are marked with an icon while the remaining vari ables are marked with an icon to distinguish the former as endogenous from the latter which are exogenous Model Equations Equations View Return to the equations view The left column shows the beginning of the equation The column on the right shows how EViews is planning on solving the model income is a function of consumption investment and government spending consumption is a function of income and consumption well Teminology hint Exogenous variables are determined outside the model and their val ues are not affected by the model s solution Endogenous variables are determined by the solution of the model Think of exogenous variables as model inputs and endoge nous variables as model outputs EViews Illustrated book Page 372 Monday February 25 2013 10 06 AM More Model Information 373 lagged consumption actually Equation Properties Double click in the view on to bring up the Properties dialog As you can see the model has pulled in the esti mated coefficients as well as the estimated standard error of the regression We re looking at a live link If we re estimated the consumption equa tion the new estimates would automatically replace the current esti mates in the model Hint How come the equation shows only CONS I G What happened to exports and imports not to
386. ture Adding dummies into the regression is easier than restructuring a workfile Hint All else equal choose the dimension with the fewest categories as the one to be handled manually EViews Illustrated book Page 287 Monday February 25 2013 10 06 AM 288 Chapter 11 Panel What s My Line There is one circumstance under which you should almost certainly use panel features rather than including dummies manually If you have lots and lots of categories panel esti mation of fixed effects is much faster Internally panel estimation uses a technique called sweeping out the dummies to factor out the dummies before running the regression dras tically reducing computational issues The time required to compute a linear regression is quadratic in the number of variables Additionally when the number of dummy variables reaches into the hundreds EViews will sometimes produce regression results using panel estimation for equations in cases in which computation using manually entered dummies breaks down Manual dummies How To The easy way to include a large number of dummies is through use of the expand func tion expand was discussed in Chapter 4 Data The Transformational Experience so here s a quick review Add to the least squares command expand cross_section_identifier droplast where the option droplast drops one dummy to avoid the dummy variable trap For example ls lnwage c ed age asian expand
387. two equations while the former doesn t The Residuals Correlation Matrix view shows the estimated cross equation correlation In this case there is very little correlation that s why the SUR estimates came out about the same the esti mates from equation by equation least squares Because coefficient estimates from all the equa tions are made jointly cross equation hypotheses are easily tested For example to check the hypothesis that the coefficients on trend are equal for cash in the hands of the public and vault cash choose View Coefficient Diagnos tics Wald Coefficient Tests and fill out the Wald Test dialog in the usual way EViews Illustrated book Page 360 Monday February 25 2013 10 06 AM Vector Autoregressions VAR 361 The answer which is unsurprising given the reported coefficients and standard errors is No the coefficients are not equal Vector Autoregressions VAR In Chapter 8 Forecasting we discussed predictions based on ARMA and ARIMA models This kind of forecasting generalizes at least in the case of autoregressive models to multiple dependent variables through the use of vector autoregressions or VARs While VARs can be quite sophisticated see the User s Guide at its heart a VAR simply takes a list of series and regresses each on its own past values as well as lags of all the other series in the list Create a VAR object either through the Object menu or the var command The
388. u and then choosing the desired template from the Apply template page Point Me The Way If you d like to point out a certain observation in your data you might want to select Draw Arrow from the right click menu or the Proc button The mouse cursor will turn into a crosshair Click at the starting point and while continuing to hold down the mouse button drag and release at the arrow endpoint You don t have to be too careful about how you initially draw your arrow because EViews allows you to change its size and position afterward Whenever you move your mouse over the arrow EViews will change the cursor to indicate what action will be taken if you drag When the cursor is a crosshair you can drag and relocate the arrow If your mouse is over the ends of the arrow the cursor indicates it will resize the arrow if you drag You can then freely drag the end point in any direction resizing and reshaping the arrow as you please Work around hint Since predefined templates don t include line shades you can t just add the Recessions graph to the Predefined template list and have recession shading globally available Hence the hint on page 174 about copy and pasting a graph that you wish to use as a template EViews Illustrated book Page 177 Monday February 25 2013 10 06 AM 178 Chapter 6 Intimacy With Graphic Objects Double click on the arrow to bring up the customiza tion dialog You can select different endpoints
389. u to choose Auto Update Options Selecting Off means that the frozen graph acts exactly like a frozen table it is a snapshot of the current graph that is disconnected from the original object If you choose Manual or Automatic EViews will create a frozen chilled graph snapshot of the current graphical output that can update itself when the data in the origi nal object changes Every frozen object is kept either as a table shown in the workfile window with the icon or a graph shown with the icon No matter how an object began life once frozen you can adjust its appearance by using custom ization tools for tables or graphs Hint Any untitled EViews object will disappear when you close its window If you want to keep something you ve frozen use the button EViews Illustrated book Page 394 Monday February 25 2013 10 06 AM A Comment On Tables 395 A Comment On Tables Every object has a label view pro viding a place to enter remarks about the object Tables take this a step further You can add a com ment to any cell in a table Select a cell and choose Proc Insert Edit Comment or right click on a cell to bring up the context menu Cells with com ments are marked with little red triangles in the upper right hand corner When the mouse passes over the cell the com ment is displayed in a note box EViews Illustrated book Page 395 Monday February 25 2013 10 06 AM 396 Chapter 17 Odds and
390. uary 25 2013 10 06 AM 318 Chapter 13 Serial Correlation Friend or Foe negative residuals Lightly dashed lines mark out standard error bands around zero to give a sense of the scaling of the residuals It s useful to see the actual fitted and residual values plotted together but it s sometime also useful to con centrate on the residuals alone Pick Residual Graph from the Actual Fitted Residual menu for this view In this example there are long runs of positive residuals and long runs of negative residuals providing strong visual evidence of serial cor relation Another way to get a visual check is with a scatterplot of residuals against lagged residuals In the plot to the right we see that the lagged residual is quite a good predictor of the current residual another very strong indicator of serial correlation The Correlogram Another visual approach to checking for serial correlation is to look directly at the empirical pattern of correlations between residuals and their own past values We can com pute the correlation between and the correlation between and and so on Since the correlations are of the residual series with its lagged self these are called autocorrelations If there is no serial correlation then all the correlations should be approximately zero although the reported values will differ from zero due to estimation error Hint For the first order serial correlation model that ope
391. umb when cells are expected to have fewer than five observations the use of test statistics gets a little dicey In such cases EViews prints a warning mes sage as it s done here x2 x2 EViews Illustrated book Page 219 Monday February 25 2013 10 06 AM 220 Chapter 7 Look At Your Data The first table shows counts for grades and test scores for non resi dents For example there were four out of state residents with top grades and low test scores The second table gives the same information for Wash ington residents Together these tables describe the joint distribution of grades and test scores conditional on residency The third table gives the joint distri bution of grades and test scores unconditionally with respect to resi dency In other words it s the same two way table we saw before You can see that with lots of catego ries an N Way tabulation can be really really long If our last series had been 50 states instead of just yes no for Washington we d have gotten 50 conditional tables and one unconditional table You can imagine how much output there would be with a fourth or fifth variable in the cross tabulation Does the order in which the series appear in the group matter The answer is no and yes Whatever order you specify you get all the possible conditional cell counts Since you get the same information regardless of the order specified there s a sense in which the order is irr
392. ur sample includes a break EViews will give an error message In this example we had to use a subset of our data to accommodate the continuous sample requirement EViews Illustrated book Page 352 Monday February 25 2013 10 06 AM ARCH etc 353 ARCH coefficients appear below the structural coefficients The ARCH coefficient our estimated is both large 0 75 and statistically sig nificant In addition to the usual results the View menu offers Garch Graph Garch Graph provides a plot of the predicted conditional variance or the conditional standard deviation Selecting Garch Graph Conditional Variance we see that higher vari ances occur early and late in the sample plus an enormous spike in 1952 Perhaps the variance spike is really there or perhaps ARCH 1 isn t the best model EViews offers a wide choice from the extended ARCH family The Model dropdown offers four broad choices Each broad choice is further refined with various options Most simply you can specify the order of the ARCH or GARCH Generalized ARCH model in the dialog fields just below Model g1 t 8 4 EViews Illustrated book Page 353 Monday February 25 2013 10 06 AM 354 Chapter 14 A Taste of Advanced Estimation You can also choose from a variety of error distributions using the Error distribution dropdown One of the most interesting applications of ARCH is to put the time varying variance back into the structural equation
393. used in gen erating the data Now we can run the regression in deviations from means and not incidentally illustrate the use of auto series ls y mean y x mean x EViews Illustrated book Page 93 Monday February 25 2013 10 06 AM 94 Chapter 4 Data The Transformational Experience Note that the output from this sec ond regression is identical to the first demonstrating the theorem but not having anything particular to do with EViews You ll also see that EViews has expanded mean y to mean y 1 100 We ll explain this expansion of the mean function in just a bit FRML and Named Auto Series Have a calculation that needs to be regularly redone as new data comes in perhaps a calcu lation you do each month on freshly loaded data Define a named auto series for the calcula tion When you load the fresh data the named auto series automatically reflects the new data Named auto series are a cross between expressions used in a command y mean y and regular series series yyy y mean y Since it s simple we ll show you how to cre ate a named auto series and then talk about a couple of places where they re particularly useful Creation of a named auto series is identical to the creation of an ordinary series except that you use the command frml rather than series To create a named auto series equal to y mean y give the command frml y_less_mean y mean y The auto series Y_LESS_MEAN is
394. used to check for the January effect historically U S stocks per formed unusually well in January could be coded series january month 1 The command show january weekday 5 tells us both about January and about Fridays as shown to the right If In place of the if statement of many program ming languages EViews has the recode s x y function If S is true for a par ticular observation the value of recode is X otherwise the value is Y For example an alterna tive to the method presented earlier for choosing the larger observation between two series is series bigger recode one_2_3 gt two_3_1 one_2_3 two_3_1 a b a x m jx N m j2 a b m j EViews Illustrated book Page 91 Monday February 25 2013 10 06 AM 92 Chapter 4 Data The Transformational Experience Not Available Functions Ordinarily any operation involving the value NA gives the result NA Sometimes particularly in making comparisons this leads to unanticipated results For example you might think the compari son x 1 is true if X equals 1 and false otherwise Nope As the example shows if X is NA then x 1 is not false it s NA EViews includes a set of special functions to help out with handling NAs notably isna X eqna X Y and nan X Y isna X is true if X is NA and false otherwise eqna X Y is true if X equals Y including NA values nan X Y returns X unless X
395. ust as easily told the story above for a variable that was constant over time while varying across countries This would lead to a period fixed effect EViews panel fea tures allow for cross section fixed effects period fixed effects or both Setting Up Panel Data The easiest way to set up a panel workfile is to start with a nonpanel workfile in which one series identifies the period and one series identifies the cross section The file PWT61Extract wf1 has information on both real GDP relative to the United States and on population for a large number of countries for half a century It also contains a series ISO CODE that holds an abbreviation for each country and a series YR for the year Hint The subscript on z is just i not it as a reminder that z varies across countries but not time Hint If two dimensions can be used rather than one why not three dimensions rather than two Why not four dimensions EViews only provides built in statistical support for two dimensional panels In the section Fixed Effects With and Without the Social Contrivance of Panel Structure below you ll learn a technique for handling fixed effects without creating a panel structure The same technique can be used for estimat ing fixed effects in third and higher dimensions yit a bxit gzi uit ai ai a gzi yit ai bxit uit ai EViews Illustrated book Page 276 Monday February 25 2013 10 06 AM Se
396. ustrated book Page 389 Monday February 25 2013 10 06 AM 390 Chapter 16 Get With the Program 4 The end result is a series of estimation results one for each repetition of step 2 We can then characterize the empirical distribution of these results by tabulating the sam ple moments or by plotting the histogram or kernel density estimate The following program sets up space to hold both the simulated data and the estimation results from the simulated data Then it runs many regressions and plots a kernel density for the estimated coefficients store monte carlo results in a series checked 4 1 2004 set workfile range to number of monte carlo replications wfcreate mcarlo u 1 100 create data series for x NOTE x is fixed in repeated samples only first 10 observations are used remaining 90 obs missing series x x fill 80 100 120 140 160 180 200 220 240 260 set true parameter values beta1 2 5 beta2 0 5 set seed for random number generator rndseed 123456 assign number of replications to a control variable reps 100 begin loop for i 1 to reps set sample to estimation sample smpl 1 10 simulate y data only for 10 obs series y beta1 beta2 x 3 nrnd regress y on a constant and x equation eq1 ls y c x set sample to one observation smpl i i and store each coefficient estimate in a series series b1 eq1 coefs 1 series b2 eq1 coefs 2 EViews Illustrated
397. uto series play in making forecasts Simple Sample Says As you ve no doubt gathered by now the statement Operations are performed on an entire series at a time is a hair short of being true The fuller version is Operations are performed on all the elements of a series included in the current sample A sample is an EViews object which specifies which observations are to be included in oper ations Effectively you can think of a sample as a list of dates in a dated workfile or a list of observation numbers in an undated workfile Samples are used for operational control in two different places The primary sample is the workfile sample This sample provides the default control for all operations on series by telling which observations to include Specific commands occasionally allow specification of a secondary sample which over rides the workfile sample When a workfile is first created the sample includes all observa tions in the workfile The cur rent sample is shown in the upper pane of the workfile win dow In this example the work file consists of five daily observations beginning on Mon day January 3 2005 and ending on Friday January 7 2005 Here s the key concept in specifying samples Hint frml is a contraction of formula EViews Illustrated book Page 95 Monday February 25 2013 10 06 AM 96 Chapter 4 Data The Transformational Experience Samples are specified as one or more p
398. uto series and then estimate our forecast ing equation frml currgrowth 1200 dlog curr ls currgrowth trend currgrowth 1 expand month EViews Illustrated book Page 236 Monday February 25 2013 10 06 AM Forecasting Beneath the Surface 237 When we hit in the equation window we get a Forecast dialog that looks just a little different The default choice in the dropdown menu is Ignore formulae within series which means to forecast the auto series currency growth in this case The alternative choice is Substitute formulae within series Choose this option and you re offered the choice of forecasting either the underlying series or the auto series We ve cho sen to forecast the level of currency and set the forecast sample to match the forecast sample in the previous example Hint If you use an expression for example 1200 dlog curr rather than a named auto series as the dependent variable you get pretty much the same choices although the dialogs look a little different EViews Illustrated book Page 237 Monday February 25 2013 10 06 AM 238 Chapter 8 Forecasting Our currency forecast is shown to the right Note that the forecast turns out to be a little too high Quick Review Forecasting EViews is in charge of the mechanics of forecasting you re in charge of figuring out a good model You need to think a little about in sample versus out of sample forecasts and dynamic versus static fo
399. values stored in RESID Resid Hint 2 EViews replaces the values in RESID with new residuals after each esti mation If you want to keep a set copy them into a new series as in series rememberresids resid before estimating anything else Resid Hint 3 You can store the residuals from an equation in a series with any name you like by using Proc Make Residual Series from the equation window EViews Illustrated book Page 81 Monday February 25 2013 10 06 AM 82 Chapter 3 Getting the Most from Least Squares EViews Illustrated book Page 82 Monday February 25 2013 10 06 AM Chapter 4 Data The Transformational Experience It s quite common to spend the greater part of a research project manipulating data even though the exciting part is estimating models testing hypotheses and forecasting into the future In EViews the basics of data transformation are quite simple We begin this chapter with a look at standard algebraic manipulations Then we take a look at the different kinds of data numeric alphabetic etc that EViews understands The chapter concludes with a look at some of EViews more exotic data transformation functions Your Basic Elementary Algebra The basics of data transformation in EViews can be learned in about two seconds Suppose we have a series named ONE_SECOND that measures in microseconds the length of one second You can download the workfile BasicAlgebra wf1 from the EViews we
400. vari able To carry out the forecast we would need to know the November 2004 inflation rate But suppose we were forecasting for November 2104 There s no way we re going to be able to plug in the right value of inflation 100 years from now No inflation rate no forecast The cardinal rule of practical forecasting is Table 3 x Estimated Parameters Value of x Product TREND 0 001784 1047 1 87 0 469423 3 4105 1 60 MONTH 11 8 321732 1 0 8 32 11 79 12 42 forecast error 0 63 y t a b xt x x x G2004m10 G 2004m11 G2004m11 x x EViews Illustrated book Page 226 Monday February 25 2013 10 06 AM Theory of Forecasting 227 Know the values of the explanatory variables during the forecast period or know a way to forecast the required explanatory variables We ll return to know a way to forecast the required explanatory variables in the next sec tion Knowing that you will have to deal with the cardinal rule in Step 3 sometimes influences what you do in Step 1 The corollary to the practical rule of forecasting is There s not much point in developing a great model for forecasting if you won t be able to carry out the forecast because you don t know the required future values of all of the explanatory variables This aspect of model development is important but doesn t really have anything to do with using EViews so we ll leave it at that
401. verage LSAT taker Tests By Classification We know from our work earlier in the chapter that out of state applicants have a slightly higher average GPA than do in state applicants Is the difference statistically significant Open the GPA series and use the menu Equality Tests by Classification to get to the Tests By Classification dialog We ve filled out the Series Group for classify field with the series WASH since that s the classifying variable of interest Buzzword hint We d say that the average UW applicant LSAT is statistically signifi cantly different from the average of all test takers EViews Illustrated book Page 212 Monday February 25 2013 10 06 AM Tests On Series 213 Since there are only two categories in this problem in state and out we need only look at the reported t statis tic and its associated p value If there were more than two categories we would have to rely on the F statistic with exactly two categories the F is redundant with the t While the difference between in state and out of state GPAs is very small statistically it s highly significant Time series tests Five tests of the time series properties of a series appear in the View menu Exploring these tests would take us too far afield for now but Correlograms and Unit Root Tests will be dis cussed in Chapter 13 Serial Correlation Friend or Foe Hint Finding a difference that is statistically significant but
402. view 413 ual graph so they ve effectively already been discussed at length in Options Options Options in Chapter 6 Intimacy With Graphic Objects One of the pages Exporting doesn t appear as an option for an individual graph but we discussed this tab in the same chapter in The Impact of Globalization on Intimate Graphic Activity The default format for saving graphics is Enhanced Meta file emf EMF is almost always the best choice How ever you may want to switch to Encapsulated PostScript eps if you send output to very high resolution devices If you re a LaTeX user you may also find eps files easier to deal with You may also save files to Graphics Inter change Format gif Portable Network Graphics png Joint Photographic Experts Group jpg Portable Document Format pdf and Bitmap bmp files GIF and PNG files are particularly useful if you wish to include graphs in web pages Quick Review If fine tuning doesn t ring your chimes you can safely avoid the Options menu entirely On the other hand if you re regularly resetting an option for a particular operation the Options menu will let you reset the option once and for all EViews Illustrated book Page 413 Monday February 25 2013 10 06 AM 414 Chapter 18 Optional Ending EViews Illustrated book Page 414 Monday February 25 2013 10 06 AM Index Symbols exclamation point 384 caret symbol 228 question mark in s
403. wage log EViews Illustrated book Page 111 Monday February 25 2013 10 06 AM 112 Chapter 4 Data The Transformational Experience Many To One Mappings Value maps can be used to group a range of codes for the purpose of display Instead of a sin gle value in the value map enter a range in parentheses For example inf 12 specifies all values less than 12 Parentheses are used to specify open intervals square brackets are used for closed intervals So inf 12 is all values less than or equal to 12 The series ED in CPSMAR2004Extract wf1 measures education in years We could use the value map shown to the right to group educa tion into three displayed values The many to one mapping is only for display purposes If we analyze the data all the underlying categories are still there For example here s a tabula tion of ED Everyone with less than a high school education is labeled drop out but they re still tabulated in sepa rate categories according to the years of education they ve had That s why we see seven dropout rows on the right Relative Exotica EViews has lots of functions for transforming data You ll never need most of these func tions but the one you do need you ll need bad Stats By We met several data summary functions such as mean above in the section Functions Are Where It s Sometimes one wants a summary statistic computed by group We might w
404. web www eviews com February 25 2013 Editors Meredith Startz and IHS Global Inc EViews Illustrated book Page 2 Monday February 25 2013 10 06 AM EViews Illustrated is dedicated to my students of many years especially those who thrive on organized chaos and even more to those who don t like chaos at all but who nonetheless manage to learn a lot and have fun anyway Acknowledgements First off I d like to thank the entire EViews crew at IHS EViews for their many suggestions and you d like to thank them for their careful review of the manuscript of this book Next thank you to David for letting me kibitz over the decades as he s built EViews into the leading econometric software package Related thanks to Carolyn for letting me absorb much of David s time and even more for sharing some of her time with me Most of all I d like to thank my 21 year old editor daughter Meredith She s the world s best editor and her editing is the least important contribution that she makes to my life EViews Illustrated book Page 3 Monday February 25 2013 10 06 AM EViews Illustrated book Page 4 Monday February 25 2013 10 06 AM Table of Contents FOREWORD 1 CHAPTER 1 A QUICK WALK THROUGH 3 Workfi
405. wed how to report results both inline and in a display table In both cases standard errors appear in parentheses below the associated coefficient estimates Standard errors in parentheses is really the first of two and a half reporting conventions used in the statistical literature The second conven tion places the t statistics in the parentheses instead of standard errors For example we could have reported the results from EViews inline as Both conventions are in wide use There s no way for the reader to know which one you re using so you have to tell them Include a comment or footnote Standard errors in paren theses or t statistics in parentheses Fifty percent of economists report standard errors and fifty percent report t statistics The remainder report p values which is the final convention you ll want to know about Where Did This Output Come From Again The top panel of regression output shown on the right sum marizes the setting for the regression The last line Included observations is obviously useful It tells you how much data you have And the next to last line identifies the sample to remind you which observations you re using Hint EViews automatically excludes all observations in which any variable in the specification is NA not available The technical term for this exclusion rule is list wise deletion volumet log 2 629649 29 35656 0 017278
406. when forecasting for real Nomenclature hint This is sometimes called in sample forecasting see Chapter 8 Forecasting 0 8365 0 8365 0 83652 EViews Illustrated book Page 328 Monday February 25 2013 10 06 AM ARMA and ARIMA Models 329 Using our AR 1 model we ve constructed three forecasts dynamic static and structural ignore ARMA this last forecast leaving out the contribution of the ARMA terms entirely The static and dynamic forecasts are identical for the first period they re supposed to be of course In general the static forecast tracks the actual data best followed by the dynamic forecast with the structural fore cast coming in last ARMA and ARIMA Models The basic approach of regression analysis is to model the dependent variable as a function of the independent variables The addition of ARMA errors augments the regression model with additional information about the persistence of errors over time A widely used alterna tive variously called Time Series Analysis or Box Jenkens analysis or ARMA or ARIMA modeling directly models the persistence of the dependent variable Estimation of ARMA or ARIMA models in EViews is very easy We begin with a short digression into the unit root problem and then work through a pure time series model of NYSE volume Who Put the I in ARIMA Series that explode over time can be statistically problem atic Most of statistical theory requires th
407. work around this for matting information by looking for data presented using the HTML table format If an HTML file doesn t read smoothly it s likely that the data has been formatted to look nice when displayed but that the table format wasn t used Reading Is Funkadelic Clever as EViews is at interpreting data it s not as smart as you are We ve seen that the dia logs include a large number of manual customization features We ve discussed some cases where automatic recognition doesn t work Here s a more inclusive but by no means exhaustive list of issues Most of the time you can use the customization features to read data with these problems 1 Multi line observations 2 Streamed observations 3 Fixed width but undelimited data 4 Dates split across multiple columns for example month in one column and year in another Hint Sometimes the Open dialog remains open for what seems like a long time while EViews processes the data from the web Be patient EViews Illustrated book Page 50 Monday February 25 2013 10 06 AM Adding Data To An Existing Workfile Or Being Rectangular Doesn t Mean Being Inflexible 51 5 Multiple tables in one file 6 Data in which one space is not a delimiter but multiple spaces are 7 Header lines that are interpreted as data Here are a couple of problems that you generally can t fix in the read dialog 1 Variable length alphabetic data recognizable
408. xed type 143 model solution 370 371 multiple 164 165 overlapping lines 141 142 panel data 283 284 pie 153 155 printing 124 quantile plots 207 quantiles 207 residual 304 305 317 318 rotating 155 saving 126 scale 124 140 141 scatter plots 13 146 147 seasonal 134 136 shading 171 172 175 spike 134 stacked graphs 140 survivor 207 survivor plots 207 templates 174 177 type selecting 181 views selecting 131 XY line 149 153 group views 119 grouping classification and 202 panels 285 287 pools 303 series 30 32 214 221 H Help EViews Forum 1 help resources 401 heteroskedasticity 306 308 338 340 351 355 high to low frequency conversion 249 250 histograms 5 197 199 203 207 History field 30 HTML data importing 50 hypothesis testing 67 68 74 78 211 212 I id series 32 36 identifiers 255 if in sample specification 97 98 if statements 91 implicit equations 380 importing from clipboard 48 241 from Excel 40 43 graphics files 126 methods 37 38 39 into pages 241 pooled 309 311 from other file formats 47 from text files 43 47 from web 48 51 impulse response 362 363 independent variables 14 63 65 66 ini files 412 in sample forecasting 227 interest rates 117 Internet importing data from 48 51 ISNA function 92 J Jarque Bera statistic 199 JPEG 126 K kernel density 206 keyboard focus 407 Keynesian Phillips curve 345
409. y log x2 EViews Illustrated book Page 380 Monday February 25 2013 10 06 AM Chapter 16 Get With the Program EViews comes with a built in programming language which allows for very powerful and sophisticated programs Because the language is very high level it s ideal for automating tasks in EViews I Want To Do It Over and Over Again If you have repetitive tasks create an EViews program and run it as needed EViews pro grams are not objects stored in the workfile You don t make them with Object New Object An EViews program is held either in a program window created with File New Program or in a text file on disk ending with the extension prg The program window at the right holds three standard commands Every time you click these three commands are executed Hint Because EViews programming language is very high level it s not very efficient for the kind of tasks which might be coded in C or Java Hint To become a proficient author you read great literature In the same vein to become a skilled EViews programmer you should read EViews programs EViews ships with a variety of sample programs You can find them under Help Quick Help Reference Sample Programs amp Data Appendix Sample Programs at the end of these chapters includes several of the sample programs with annotations Hint Clicking the button in a program window saves the file to disk with the extension prg But you re free to cr
410. y type in the EViews command pane ls logvol c t as shown below y a b t a b t 400 b b 4 2 5 400 0 0 01625 a b EViews Illustrated book Page 13 Monday February 25 2013 10 06 AM 14 Chapter 1 A Quick Walk Through In EViews you specify a regression with the ls command followed by a list of variables LS is the name for the EViews command to esti mate an ordinary Least Squares regression The first variable is the dependent variable the variable we d like to explain LOGVOL in this case The rest of the list gives the independent variables which are used to predict the dependent variable Whoa a minute LOGVOL is the variable we created with the logarithm of volume and T is the variable we created with a time trend But where does the C in the command come from C is a special keyword signaling EViews to estimate an intercept The coeffi cient on the variable C is just as the coefficient on the variable T is Hint Sometimes the dependent variable is called the left hand side variable and the independent variables are called the right hand side variables The terminology reflects the convention that the dependent variable is written to the left of the equal sign and the independent variables appear to the right as for example in volumet log a bt
411. y hint for graphing multiple series Individual series in a Group window may have NAs for different observations As a result distribution graphs may be drawn for different samples For example the graph to the right makes it appear that 3 month Treasury rates are much more likely than are 1 year rates to be nearly zero In fact what s going on is that our sample includes 3 month but not 1 year rates from the Great Depression Looking at a line graph with the same histograms on the axis bor der it becomes evident that the 1 year rates enter our sample at a later date EViews Illustrated book Page 205 Monday February 25 2013 10 06 AM 206 Chapter 7 Look At Your Data Kernel Density Graphs A kernel density graph is in essence a smoothed histogram Often a histogram looks choppy because the number of observations in a given bin is subject to random variation This is particularly true when there are relatively few observations The kernel density graph smooths the variation between nearby bins The User s Guide describes the various options for controlling the smoothing Using the default choices gives a nice picture for our GPA data There s no law about the best way to accomplish this smooth ing but the default frequently works well We see again that applicant grades are concen trated around 3 4 or 3 5 and that there is a long lower tail Theoretical Distribution EViews will fit any of a number of theoretical prob
412. you froze it Save graph to disk shows up on the right click menu EViews Illustrated book Page 396 Monday February 25 2013 10 06 AM Unsubtle Redirection 397 Unsubtle Redirection Inside Output You hit and output goes out right Not necessarily as EViews provides an option that allows you to redirect printer output to an EViews spool object The spool object collects output that would otherwise go to the printer This gives you an editable record of the work you ve done The window below is scrolled so that you can see part of a table and part of a graph each of which had previously been redirected to this spool Objects are listed by name in the left pane and object contents are displayed in the right pane EViews Illustrated book Page 397 Monday February 25 2013 10 06 AM 398 Chapter 17 Odds and Ends Outside Output You hit and output at least out put that doesn t stay inside goes to the printer right Not necessarily as EViews provides an option that allows you to redirect printer output to a disk file instead Choosing the Redirect radio button in the Print dialog leads to output destination choices RTF file adds the output to the end of the speci fied file in a format that is easily read by word processing programs Text file writes text output as a standard unfor matted text file and sends graphic out put to the printer Frozen objects freezes the object that you said to print
413. ystrokes easier than Save As But while Save protects you from com puter failure it doesn t substitute for an Undo feature Instead it copies the current workfile in memory including all the changes you ve made on top of the version stored on disk Pedantic note EViews does have an Undo item in the usual place on the Edit menu It works when you re typing text It doesn t Undo changes to the workfile EViews Illustrated book Page 19 Monday February 25 2013 10 06 AM 20 Chapter 1 A Quick Walk Through Forecasting We have a regression equation that gives a good explanation of Let s use this equation to forecast NYSE volume Hit the button on the equation win dow to open the forecast dialog Notice that we have a choice of fore casting either volume or When you use a function as a dependent variable EViews offers the choice of forecasting either the function or the underlying variable The one we actually care about is volume taking logs was just a trick to get a better statistical model Leave the dialog set to forecast vol ume Uncheck Forecast graph Fore cast evaluation and Insert actuals for out of sample observations In the Forecast sample field enter 2001q1 2004q1 Your dialog should look something like the one shown volume log volume log EViews Illustrated book Page 20 Monday February 25 2013 10 06 AM What s Ahead 21 EViews creates a series of fore c
414. zation in Chapter 5 Picture This Fill Areas The Fill Areas page does the same job for filled in areas in bar graphs for example that the Lines amp Symbols page does for lines In addition the Bar Area Pie page is the place to control label ing outlining and spacing of filled areas In Distinguishing Factors in the previous chapter we used Within graph category identifica tion to have EViews automatically select visually distinct colors for different graph elements EViews automatic selection produced the graph below Two rules to remember Multiple colors are much better than patterns in helping the viewer distinguish different series in a graph Multiple colors are not so great if the graph is printed in black and white EViews Illustrated book Page 189 Monday February 25 2013 10 06 AM 190 Chapter 6 Intimacy With Graphic Objects Using the Fill Areas page to set hatching for the first series the dialog says series even though the bars are really categories of a single series produces the more visually distinc tive version shown here EViews Illustrated book Page 190 Monday February 25 2013 10 06 AM Options Options Options 191 Boxplots The BoxPlots page offers lots of options for deciding which elements to include in your boxplot as well as color and other appear ance controls for these elements For a light review of the various elements se

EViews Illustrated for Version 7

Contents

Download Pdf Manuals

Related Search

Related Contents