Home

Appendices - Time Series Modelling (TSM)

Contents

1. decl dStat iDist iDFl iDF2 return dStat iDist iDFl iDF2 22 James Davidson 2015 The format of the return value is the same as for UserTest A single statistic or multiple statistics can be returned User supplied critical values are implemented in the same way as described above Notes 1 Either a single name or array of names can be returned from the function In the case that m statistics are returned the array should be of dimension m 1 The first element is a heading for the test group followed by names for the individual statistics Even if only one statistic is returned the first of a pair of names is used to provide a heading and the second appears on the line before the statistic value is printed Returned names are ignored by TSM unless bMode 0 so conditionally bypassing the naming steps can save CPU time 2 Parameters can be passed to the supplied function by naming them in the Parameter Names fields and assigning values in the Values Equations dialog Note that these values are simply passed through to the function They are not changed by the program and can be arbitrary for example integer values to select from a list of test options 3 The supplied function can call TSM program functions as described in the document TSM4 Programming Reference t smod4prg pdf These can be used to change model settings compute and access estimates and other results set starting v
2. 19 Returning Test INAIN so aes du dee qua td 21 Passing Settings to the Function 6 4 21 Critical Values e 21 Coding a Test with Hessian Contributions sese 22 Coimputifig 3 en oreet RE or p TO npe e E OR Lee eus 22 Generating Data cides edes 23 Controlling Program Functions ui e repe 24 Loading The COG asc teilte 25 Maintaining a Function 1 1 1 0 seien eene 26 Exporting Ox Code isi uas a ra AAT A Mx aua 27 Documenting the Code Eo Se ER qo aput ao a qu d ahs 28 Debugging Code Running in Diagnostic Mode sese 28 Appendix D Calling the Code from an Ox Program 30 Appendix E Saving and Loading Batch Settings in the GUI 3 Appendix F Using TSM in the Classroom esee 33 Installation and USC aos eo oS EUER DIN LE TO e oe 33 Linear Regression Mode uiii cene ite 33 Simplified sooo b tice td etta bs E 33 1 James Davidson 2015 Easy Equation Graphics Display o aq q
3. 41 James Davidson 2015 IV UPPER TAIL CRITICAL VALUES FOR amp 1 Kwiatkowski al 1992 Table I 1 Nu Upper tail percentiles of the distribution of V r dr Critical level 10 05 025 01 Critical value 0 347 0 463 0 574 0 739 1 1 Upper tail percentiles of the distribution of I V r dr Critical level 10 05 025 01 Critical value 0 119 0 146 0 176 0 216 V NYBLOM HANSEN TEST CRITICAL VALUES from Hansen 1990 P Lc gt v 12 all 075 05 025 01 Parameters 1 243 292 398 470 593 1 748 2 469 610 670 749 898 1 07 3 679 846 913 1 01 1 16 1 35 4 883 1 07 1 14 1 24 1 39 1 60 5 1 08 1 28 1 36 147 1 63 1 88 6 1 28 1 49 1 58 1 68 1 89 2 7 1 46 1 69 1 78 1 90 2 10 2 35 8 1 66 1 89 1 99 2 11 2 33 2 59 9 1 85 2 10 249 2 32 2 55 2 82 10 2 03 2 29 2 40 2 54 2 76 3 05 11 2 22 2 49 2 60 2 75 2 99 3 27 12 2 41 2 69 2 81 2 96 3 18 3 51 13 2 59 2 89 3 00 3 15 3 39 3 69 14 2 77 3 08 3 19 3 34 3 60 3 90 15 2 95 3 26 3 38 3 54 3 81 4 07 16 3 14 3 46 3 58 3 75 4 01 4 30 17 3 32 3 64 3 77 3 95 4 21 4 51 18 3 50 3 83 3 96 4 14 4 40 4 73 19 3 69 4 03 4 16 4 33 4 60 4 92 20 3 86 4 22 4 36 4 52 4 79 5 13 42 James Davidson 2015 VI F CRITICAL VALUES from Kiefer and Vogelsang 2002b P F lt v 90 95 97 5 99 Restrictions 1 14 28 23 14 33 64 51 05 2 17 99 26 19 35 56 48 74 3 21 13 29 08 37 88 51 04 4 24 24 32 42 40 57
4. 10 5 1 10 5 1 10 5 1 0 5 1 19 81 22 36 27 69 21 06 23 68 29 14 22 31 25 30 58 0 49 1 08 21 05 23 73 29 67 22 52 25 17 31 23 24 3 2693 3273 048 1 17 21 87 2453 30 23 2345 26 11 32 11 25 13 27 82 33 55 0 47 0 27 2246 2524 30 87 24 00 26 75 32 57 25 65 28 42 34 08 0 45 1 49 23 31 25 95 31 62 24 98 27 65 33 39 26 59 29 25 35 16 0 4 2 25 24 91 27 56 33 38 2646 292 34 97 2821 31 07 36 59 0 35 3 45 26 2877 34 65 2752 303 36 1 2923 3211 37 32 0 3 5 44 2695 297 35 26 2841 31 08 37 02 30 21 32 8 38 28 0 25 9 27 64 30 48 35 85 29 31 31 83 37 25 30 93 33 65 38 93 0 2 16 28 36 31 1 36 47 29 98 32 65 38 41 31 68 34 41 40 07 0 15 32 11 29 13 31 8 37 04 30 67 33 45 39 22 32 51 35 06 40 56 0 1 81 30 00 3262 37 64 31 48 3422 39 6 33 28 35 76 41 28 0 05 361 30 9 33 42 38 44 32 48 35 40 1 3425 3674 42 05 16 p 117 18 To 10 5 1 10 5 1 1096 5 1 0 5 1 23 54 26 3 32 2476 27 58 33 43 25 99 28 87 34 81 0 49 1 08 25 32 2817 34 11 26 77 29 72 33 57 27 85 30 79 36 82 0 48 1 17 26 22 2905 35 09 27 65 30 58 36 78 28 76 31 57 37 7 047 027 2683 29 73 35 87 2816 3125 37 38 29 32 32 24 38 65 0 45 1 49 27 75 30 59 36 88 2914 3218 38 25 3038 3317 39 31 0 4 2 25 29 34 32 28 38 56 3082 33 74 39 66 3211 351 40 99 0 35 3 45 3056 33 4 39 18 31 6 34 86 40 81 334 3623 41 84 0 3 5 44 31 56 3441 40 29 32 99 35 83 41 73 34 33 37 09 42 69
5. 46 James Davidson 2015
6. A variable width font is used in the results window so that the usual formatting with columns of data aligned is not obtained A work around to see the output in its usual formatting is to highlight the required text right click and choose Copy and then paste the text into Q4Wine Notepad which does offer the option of a Courier font The second issue is that DOS commands issued by Ox can fail to execute and in particular graphics are not displayed by Gnuplot automatically However the PLT files that TSM creates to be processed by Gnuplot are placed in the working directory With Gnuplot installed double clicking these icons in Q4Wine Explorer brings up the desired plots SsfPack State Space Modelling SsfPack Basic is a free for academic research and teaching purposes download from http www ssfpack com If it is installed as a package in your Ox installation its facilities can be accessed from TSM Simply extract the ssfpack folder from the download file ssfpack_basic_30 zip into the oxmetrics7 ox packages folder and then re run the TSM installation program SsfPack will be detected automatically and a dialog to use its features can be opened from the Setup menu Usage is fully documented in the TSM User s Manual in conjunction with the SsfPack documentation Please note that commercial and government users should purchase a licence for SsfPack as for the Ox package itself See the web page for details 8 James Davidson
7. Appendix H Running TSM in the Condor environment Condor is a High Throughput Computing HTC system that currently runs on a number of university networks around the world It allows a program to be launched from a Condor enabled workstation to be run on another workstation on the network that is currently idle and the outputs from the run returned to the originating machine as if it had been run locally TSM exploits the implied availability of multiple processors by allowing Monte Carlo experiments to be split into multiple small instances to be run in parallel The results from these are aggregated automatically by TSM to be presented as the results of a single large experiment The User s Manual gives details The implementation of Condor in TSM uses the TSM facility to create a free standing Ox source file with extension ox that can be compiled with the TSM code This source file contains the model specifications set up interactively by the user in TSM using TSM s scripting language and calls the various functions to run the job either an estimation run or a Monte Carlo experiment Running the Ox executable ox1 exe from the command line with this file as its argument runs the estimation or simulation job exactly as if the Run button had been pressed in the TSM GUI In the basic single run implementation the results are returned as a text file that can be loaded for viewing in the TSM results window When parallel insta
8. appear before the calling function in the ox file 6 See the Ox documentation for additional guidance on coding The rule for fast code is to use built in Ox functions as much as possible While nonlinear recursions such as the above example may be unavoidable explicit loops are slow to execute compared to the equivalent operations using Ox matrix functions 7 The bMode flag can be used to reduce computation time by storing the results of calculations that need to be repeated in each of a sequence of calls The statement UserStore x stores the contents of the object x which can be of any type Any existing contents are overwritten The statement x UserRetrieve 16 James Davidson 2015 copies to x whatever has been stored by a previous call to UserStore If nothing has been stored 0 is returned Thus suppose the user s function contains the statements decl x if bMode x UserStore x else x UserRetrieve where the ellipsis represents the code for computing x This will compute and store x on the first call and retrieve it on subsequent calls If there is more than one object to be stored and retrieved define x as an array It is even possible to share the storage space between several user functions by making sure that the array is initialized on the first call Recalling that UserRetrieve returns 0 if nothing has been stored in it consider the following example decl z x UserRetrieve
9. 52 39 5 27 81 35 97 44 78 56 92 6 30 36 38 81 47 94 60 81 7 33 39 42 08 50 81 62 27 8 36 08 45 32 54 22 67 14 9 38 94 48 14 5741 69 67 10 41 71 50 75 59 98 72 05 11 44 56 53 7 63 14 74 74 12 47 27 56 7 65 98 78 8 13 50 32 60 11 69 46 82 09 14 52 97 62 83 72 46 85 12 15 55 71 65 74 75 51 88 86 16 58 14 68 68 78 09 91 37 17 60 75 70 59 80 94 94 08 18 63 35 73 76 83 63 97 41 19 65 81 76 42 86 2 99 75 20 68 64 79 5 89 86 103 2 21 70 8 82 92 32 105 4 22 73 41 84 76 94 54 108 23 76 19 87 15 98 06 111 8 24 78 4 89 67 100 4 114 7 25 81 21 92 7 103 5 117 6 26 83 59 95 49 106 6 120 8 27 85 83 97 57 108 8 123 4 28 88 11 99 48 110 7 124 5 29 90 92 102 9 114 6 129 6 30 93 63 105 8 117 5 132 1 43 James Davidson 2015 VII STRUCTURAL CHANGE TEST CRITICAL VALUES from Andrews 1993 pel 2 pes To x 10 5 1 10 5 1 10 5 1 0 5 1 271 3 84 663 461 599 921 625 7 81 11 34 0 49 1 08 347 473 7 82 542 686 103 7 19 8 83 12 58 0 48 1 17 3 79 5 1 8 26 5 8 7 31 10 71 7 64 9 29 13 05 047 0 27 402 538 865 612 7 67 11 01 7 98 9 62 13 39 0 45 1 9 438 5 91 9 00 66 811 11 77 850 1015 14 23 04 2 25 510 657 982 745 902 12 91 946 1117 14 88 0 35 3 45 5 59 7 05 1053 806 9 67 13 53 10 16 12 05 15 71 03 544 605 7 51 10 91 8 57 10 19 14 16 10 76 12 58 16 24 0 25 9 646 7 93 1148 91 10 75 1447 1129 13 16 166 0 2 16 6 80 845 11 69 9 59 1126 15 09 11 8 13 69 1728 0 15 3211 717 8 85 12 35 1001 1179 15 51 12 27 1415 17
10. 68 0 1 81 7 60 9 31 12 69 10 5 1227 16 04 12 81 14 62 18 28 0 05 361 8 19 9 84 13 01 112 12 93 16 44 13 47 15 15 19 06 p 4 p 3 p 6 To 10 5 1 10 5 1 1096 596 1 0 5 1 7 78 949 13 28 9 24 11 07 15 09 10 64 12 59 16 81 0 49 1 08 8 93 10 63 14 64 10 39 12 28 16 34 11 81 13 74 18 32 0 48 1 17 942 11 17 15 17 10 96 12 88 16 83 12 42 14 45 19 12 047 0 27 9 82 11 63 15 91 114 13 27 17 32 12 9 14 86 19 64 0 45 1 49 10 35 1227 16 64 12 05 14 18 06 13 53 15 59 20 45 04 2 25 11 39 13 32 17 66 13 09 15 16 19 23 14 71 16 91 21 6 0 35 3 45 121 1412 18 54 13 86 15 93 19 99 15 56 17 75 22 33 03 5 44 128 1479 19 1 14 58 1648 20 67 16 32 18 46 23 06 0 25 9 13 36 15 34 19 78 1517 1725 21 39 17 19 07 23 65 0 2 16 13 82 15 84 20 24 15 63 17 88 21 9 17 56 19 64 2427 0 15 32 11 14 31 16 45 20 71 16 2 18 35 22 49 18 12 20 26 24 79 0 1 81 14 94 16 98 21 04 16 87 18 93 23 34 18 78 20 82 2521 0 05 361 15 62 17 56 21 54 17 60 19 61 24 18 19 49 21 56 25 96 7 8 pa To 10 5 1 10 5 1 10 5 1 0 5 1 12 02 14 07 18 48 13 36 15 51 20 09 14 68 16 92 21 67 0 49 1 08 13 27 15 52 19 93 1329 15 63 20 53 16 17 18 56 23 05 0 48 1 17 13 92 16 14 20 64 13 89 16 31 21 14 16 82 19 25 23 83 047 0 27 14 32 16 63 21 14 1443 16 74 21 72 17 26 19 74 24 8 0 45 1 49 14 97 17 38 22 32 15 05 17 53 22 28 18 1 20 59 25 52 0 4 2 25 1623 1841 23 35 16 26 18 7
11. Dynamic Equation dialog Use the Make Zeros and Rename commands in the Setup Data Transformation and Editing dialog to create these placeholder variables It is the user s responsibility to have the number of variables selected matching the number of columns returned Any other model specifications are ignored 2 The function argument aName passes the string entered in the Function Name field in Model Coded Function Test This can be used as with UserFunction to select one of a set of coded specifications Note the UserFunct ion and UserLikelihood options cannot be implemented at the same time as UserGenerate These options can be used in combination in a Monte Carlo experiment by creating different models to generate the data and estimate the model respectively 3 The parameter vector vP corresponds to the supplied function parameters as in UserFunction and UserSolve These can be named and assigned values in the Model Coded Function Test dialog 4 The main difference between this function and UserSolve is that is does not use residuals generated by TSM and always returns the full generated sample in a single call It is not called sequentially observation by observation If the Coded Function option is selected in Model Dynamic Equation the simulation module uses the output from UserGenerate whenever the function is defined and its return value is not equal to 0 otherwise UserSolve is called
12. Size to Small sets this function to return 1 as well as specifying small icons This is generally the best size for Windows XP and 7 systems If Icon and Font Size is set to this function returns 1 6 and the onscreen display is scaled up accordingly Get FontSize sets the points for the dialog box text Note that this setting is relative to the selected width factor The default value of 12 points should not normally need changing Get ChFontSize sets the point sizes for lists and combo box pull down menu text Note that this setting 1s relative to the selected width factor The default value is 11 and should not normally need changing Get FontStyle sets the font style for the dialog box text GraphDelay sets a delay in 100ths of a second between calling GnuPlot and processing its output The default is 200 Increase the setting if graphic files are not being written to disk correctly Get FrColor returns the colour used for the menu bar tool bar and status bar as an RGB triplet an array of three integers from 0 255 White is the colour selected for Windows 8 8 1 and 10 installations Get BgColor returns the dialog background colour similarly Get HgColor returns the colour to be used for highlighted buttons similarly Get TextAreaBG returns a background colour of the text window as an RGB triple selectable as one of the options in Options General Get NameCharac
13. TSM installation folder in Windows Explorer Typically this is C Program Files OxMetrics7 ox packages tsmod4 b Right click the batch file t smod_runsc bat and choose Create Shortcut Right click the shortcut and select Properties d Edit the Target field and add the following run parameters following the batch file name 1 The path to the Ox executable ox1 exe typically C Program Files OxMetrics7 ox bin ii The path to the TSM installation typically C Program Files OxMetrics7 ox packages tsmod4 Use the Target field in the default shortcut as a template for these entries Set the Start in directory as desired Set the Run option to Minimized 68 nm Change the icon to tsmico ico located in the TSM Installation directory h Click OK rename the shortcut and drag it to the Start Menu or desktop Any number of such shortcuts can be created to point to starter files in different folders associated with different research projects To create additional shortcuts of the TSM4 with User Code type create the shortcut to point to tsmod_runuc bat The user code module and settings file loaded then depends on which shortcut is used to start the program The use of shortcuts can be avoided by double clicking any settings file with t sm extension in Windows Explorer to start the program The folder containing this file is then automatically assigned as the Home folder Another possibilit
14. a working directory in their personal documents folder or network space It should have a red TSM icon when TSM is installed Double clicking on the file in Windows Explorer starts the program with all the stored settings The stored data are automatically written to a file in the selected spreadsheet format and model settings and outputs are likewise automatically written to files with t sd extensions Paths to file locations within the original home folder are recreated relative to the new home folder in the installed location Hence data model files and results can be organized into folders However any data stored outside of the home folder in the original location are placed in the home folder in the new location If students change the settings and create their own models these are saved automatically in the files settings tsm and t sd files for each model Full estimation outputs including generated series graphs and optionally data and displayed results can be stored as named models in the Model Manager If they wish students can also bundle and save these items for future reference using the Export feature Maintaining and Upgrading TSM A current feature of TSM is fairly frequent upgrades and new releases One day it may settle down and stop developing but that has not happened quite yet This can be a nuisance for network managers However once the program is installed upgrading is only a matter of copying co
15. as cross platform ensuring a similar appearance under Windows and Unix 6 Also selectable are two alternative sizes for the icons and fonts Under Windows XP and Windows 7 the choice Small appears satisfactory However on a Windows 8 or Windows 10 Surface tablet the display is excessively cramped and scrollbars do not display correctly In this situation or simply if you like a big bold application choose Large These settings can be refined post installation by editing the header file tsmgui4 oxh see Appendix B The installation creates registry keys that provide icons for the special program files with extensions t sm program settings and specifications the red TSM logo and t sd model specifications results and data blue TSM logo on a page Installing 64 bit TSM This requires the 64 bit version of Ox 7 which is a component of OxMetrics 7 Professional This is a commercial product that can be purchased from Timberlake Consultants Ltd There is no 64 bit version of Ox Console as yet Also needing to be installed is the 64 bit version of the Java JRE Install TSM in the same way as for the 32 bit version with the appropriate selection at Stage 3 above Note that it is possible to have both 32 bit and 64 bit Ox installed so be careful to select the correct installation directory at Stage 1 In Windows 7 this is contained in Program Files while 32 bit software is installed in Program Files x86 Starting and
16. bundling settings model specifications and data together in one file with a t sm extension Opening this file at another installation re instates all the settings exactly reproducing the original set up Ifa file usercode ox exists in the Start in directory when the settings are exported and the option Include User Code with Exported Settings is checked in the Options General dialog the currently loaded Ox code is saved similarly and re instated at the target installation This feature can be used for various purposes to distribute coded estimators or tests to other users or simply to store all the materials associated with a project including code model settings and data in one convenient location In the following assume that when the exported settings file was created the user s code was contained in a file mycode ox say and usercode ox contained just the associated 4 include directive this is the recommended setup When the settings file is subsequently opened at the target installation the following actions are performed automatically 1 Ifthe file usercode ox does not exist in the Start in directory it is created Otherwise any existing contents are commented out and the line include mycode ox is appended 2 Acopy of mycode ox is created in the same directory 3 The copy of tsmod_run ox residing in the Start in directory is edited as necessary to compile the code as described in Loading th
17. if isarray x x new array 2 if bMode t x 0 UserStore x 2 x 0 The second function using the store should contain the same lines with 2 suitably defined except that x 0 is replaced by x 1 8 Becareful to note that the pointer aName always points to a string a name when the function is called On exit it can point to a different string and in some cases to an array of strings Coding Residuals For a single equation model UserFunction should return a column vector of residuals having cEnd cStart 1 rows For a system of equations UserFunction should return a matrix with the residuals for each equation in the columns In this case there is a built in function LocUP for locating the parameters for each equation The location vP LocUP cEq cJ contains parameter cJ in equation cEq Following the Ox convention we count from Zero so the reference to the first parameter in the first equation is vP LocUP 0 0 This function will usually be called as part of a numerical optimization sequence The bMode flag is set to 0 on the initial call and 1 on all subsequent calls until estimation is completed Any function components not depending on the parameters should be evaluated and stored on the first call and retrieved in later calls Tips If the number of lags or other features of your function are to be chosen interactively code the most general form you want to consider Then simplify the funct
18. if aName 0 UserStatistic Names 0 return MyTestl vParam mcX cStart cEnd amp names else if aName 0 UserStatistic Names 1 1 return MyTest2 vParam mcX cStart cEnd amp names else PrintCall 1 Error aName 0 not found aName 0 names MyTest1 and MyTest2 are the functions containing the actual test formulae returning their output as 4 x k matrices as specified above Note that they can also return a pointer as amp names pointing to an array of names for labelling the outputs This in turn is handed back by assigning it to aName 0 see the last line of the example function 26 James Davidson 2015 Notes 1 Only one function library can be contained in each code file Maintain different files for different projects and switch between them using the Restart command as described in the previous section 2 If the list of names to be checked through is lengthy it saves a bit of CPU time in repeated calls to store the index of the required function in UserStore at the first call when bMode 0 and retrieve it for subsequent calls when bMode 1 3 There is no naming function paired with the UserSolve function because the outputs returned from this function if defined must be paired with the outputs returned from UserFunction The same names should of course label both cases Exporting Ox Code The TSM command File Settings Export creates a portable settings file
19. is set to 0 when cStart and cEnd represent the first observation to be solved and 1 for the subsequent observations of the sample The program writes the generated values into the matrix mcX in sequence For cStart gt 0 the user can assume that mcX t 1 xnum is either the actual observation on BILIN on the first call or otherwise the value returned on the previous call The forecast period can extend beyond the end of the observed data and in this case the matrix mcX is extended with zeros It 1s the user s responsibility to make sure to specify the sample and forecast periods so that valid data are being read In general only closed models without exogenous variables can be forecast properly ex ante beyond the observed period UserSolve cannot return a function name it can only receive the name as an argument It must be paired with a UserFunct ion and if this function returns a name this is passed to UserSolve as well being used in the output 18 James Davidson 2015 Coding the Likelihood Another option is to return a user coded likelihood function In this case all the modelling options in the program are bypassed and the user has freedom to specify a complete model The arguments for UserLikelihood are exactly as for UserFunction This function must return a column vector of dimension cEnd cStart 1 representing in this case the contributions to the log likelihood for each observation that is the log dens
20. located in the designated Start in directory Linux users should copy this file manually The short cut also creates a standard code file named usercode ox by default which can serve as a template for the user s own code On installation this file contains dummy do nothing versions of all the user editable functions Edit these functions as required to include your own code Note Compilable versions of all these functions whether dummy or active must be present in the code file otherwise TSM will not start Once the code exists in usercode ox compile it by opening TSM and giving the command File Restart Load User Code and choosing the option OK This command restarts TSM after first creating or editing the file t smod_run ox as necessary so that it contains at least the three lines define USER_CODE import lt packages tsmod4 tsmgui4 gt include usercode ox This file can also be edited manually in a text editor if required OxEdit recommended Once these lines are present any code included in the code file will be compiled automatically at start up so the Restart step need not be repeated However the recommended scheme is to place the actual code in yet another Ox file to be included in usercode ox In this case in Windows a code file can be loaded automatically by giving the command File Restart Load User Code and choosing the option Select Code Fil
21. mcX const cStart const cEnd const aName const bMode decl dStat iDist iDFl iDF2 return dStat iDist iDFl iDF2 The return value is in this case a 4x1 vector dStat is the computed value of the test statistic iDist is an integer code to indicate what distribution is to be used to compute the p value for the test and iDF1 and iDF2 are degrees of freedom or other test parameters If only the statistic is returned no p value will be reported The distribution codes are iDist 2 No critical values No p values are reported equivalent to 19 James Davidson 2015 H p B ob B b B b H Dist Dist Dist Dist Dist Dist Dist il il il il returning the statistic only Dist 1 Critical values returned by the function see below Chi squared with iDF1 degrees of freedom Standard normal 2 tail test Standard normal upper tail test F distribution with iDF1 and iDF2 degrees of freedom Student s t with 1DF2 degrees of freedom 2 tail test Student s t with 1DF2 degrees of freedom upper tail test Dickey Fuller distribution with 1DF1 1 to allow for trend 0 otherwise Ne we o Ne Ne Ne NOB WN EF c e Dist 7 Dickey Fuller distribution for regression residuals with iDF1 1 to allow for trend 0 otherwise and iDF2 regressors excluding trend Dist 8 KPSS distr
22. of aDst at are the diagnostic statistics for each equation Codes to access the elements of vectors aDstat eq and mStat are given in TSM4 Programming Reference pages 34 35 Elements are 0 unless the test has been specified in the input The other items in the list of arguments are the same as for UserFunct ion etc 20 James Davidson 2015 Returning Test Names The test or tests can be given a name by including in the function a statement of the general form aName 0 Test where an identifying name is substituted in quotation marks on the right hand side This is used to label the statistic in the output In the case where two or more statistics are returned supply a name for each statistic This is done by the use of an array for example aName 0 MyStat1 MyStat2 If bMode is set to 1 any returned names are ignored In a Monte Carlo experiment the names are stored following a preliminary run with bMode 0 The CPU time for the experiment may be reduced by conditionally bypassing the code to create the name strings Passing Settings to the Function Numerical values representing test settings can be passed to the function as components of the test name in a specified format and then converted using the Ox sscan function For example suppose the test name is the string TestCase a 4 b 1 passed as aName 0 Variables ia and ib can be assigned the integer values 4 and 1 respectively using the state
23. process The file t smgui4 oxh may be customized by the user See Appendix B The Setup routine copies this file to the Start in directory only if it does not already exist The existing copy is not over written nor is it deleted by the uninstall routine Setup creates a file called t smgui4bk oxh containing the current default entries for t smgui4 oxh An upgraded t smgui4 oxh is created by combining these two files preserving the existing form of all lines containing customizable settings See Appendix B page 12 for the list of keywords identifying customizable lines If there is no pre existing copy of tsmgui4 oxh in the installation directory it is copied direct from t smgui4bk oxh Copying this file manually is an easy way to restore the default settings Troubleshooting 1 Under Windows XP Vista 7 8 10 the program should install reliably provided all the installation procedures are followed correctly In particular be sure that in the TSM installation wizard the Destination Location is correctly located within the existing Ox installation folder 2 If nothing happens when you try to start the program as above check whether the DOS command window has opened and then whether the Java Runtime Environment JRE has started These components have separate task bar icons Problems with the JRE have occasionally been reported under Windows Vista and these almost certainly have to do with the misbehaviour of certain third party fire
24. the code file This can be commented out by placing it between and pairs This arrangement lets users examine both the description and the code itself without starting a text editor However this is a viewing facility only The file cannot be edited in this window Debugging Code Running in Diagnostic Mode TSM implements an error recovery feature that allows Ox error messages to be reviewed in the event of an execution error Ox prints useful diagnostic information including the line number where the crash occurred However for this feature to work the program must be able to start so it does not help in the case of compilation and linking errors Provided the standard Ox header files are included in usercode ox the code can be compiled as a free standing module Choose the option Modules Ox compile in OxEdit which will print any error messages However if there are linking errors preventing TSM from starting with usercode ox included it is necessary to run TSM in diagnostic mode to see error messages requiring some configuration by the user For widest scope of application the procedures described assume that the user does not have write permissions for the TSM Home directory 1 Redirecting batch output Right click the shortcut used to start TSM and select Properties b Edit the Target field and enter gt messages txt at the end of the line c Click OK This directs Ox messages to the
25. the user to copy settings files seamlessly between installations even when the paths to the user documents directory may be different The first element should contain the path to the documents directory of the local installation The second element or array of elements should consist of any paths that are to be replaced by the local path if encountered while reading a file The default code is return Some run settings can be modified by compiler directives These can be included in the file tsmod run ox where they must appear before the line import lt packages tsmod4 tsmgui4 gt For inclusion on a permanent basis the directives can also be placed in tsmgui4 h Note that lines of a file with ox or oxh suffix beginning with are treated as comments and ignored by the Ox compiler This is called commenting out a line Use this method rather than deleting lines to edit your files Then the old settings are easy to restore The compiler directives take the form de fine directive where directive represents one of the following capitalized names USER_CODE 12 James Davidson 2015 Signals the inclusion of user coded functions in the run file For further details see Appendix C TEXT INPUT Set this directive if program settings are to be loaded at start up in coded form as lines an external function Text Input See Appendix E for details IEXT DEFAULTS GNUA Simila
26. this together with alternative critical values for example Coding a Test with Hessian Contributions Some tests make use of the contributions to the Hessian of the log likelihood function The best known example is the information matrix test a TSM option The following special function is provided to return these values CallHessianContributions const vParam The argument should be the parameter vector exactly as passed to the UserTest function The function returns the T x p p 1 2 matrix of Hessian contributions The tth row of this matrix contains the centred numerical second derivatives for observation f arranged by rows as the upper triangle of the Hessian with indices ordered as 1 1 1 2 1 p 2 2 2 1 p p To use this feature the user code file must contain the header extern CallHessianContributions const vParam Computing a Statistic The UserStatistic function does not use estimation outputs and is intended to be called free standing No simulation or estimation can be performed in a run specifying this option Create another model to generate data for Monte Carlo experiments Select the Coded Function option in the Model Dynamic Equation dialog in the Model Coded Equation dialog and the Statistic radio button in the Supplied Ox Functions section The function format is UserStatistic const mcX const cStart const cEnd const const aName const bMode
27. usercode ox It can be changed but should also match the name appearing in Windows batch file scripts Get BatchFileName returns the name of the Windows batch file containing the TSM start up script needed to enable automatic restarts By default this 1s tsmod_runsc bat It can be changed but must match the name appearing in the Windows Start Menu and Folder Option scripts Get SettingsFileName returns the name to be used for the standard TSM settings file which is opened at start up By default the value returned is settings tsm It may be convenient to use a different name if two or more TSM installations exist Get ListingFileName returns the name to be used for the data backup file and should have a 164 extension By default the value returned is settings tsd In general use the name matching the standard settings file Get ModelFolder Get EDFFolder Get BatchFolder These functions return default paths to storage folders for files of the indicated types as subdirectories of the Start in directory the home folder The entries returned by these functions are appended to the Start in path as set during installation they should not be complete paths They are empty by default returning such that the Start in directory contains all these files ChangeUserPath returns a two element array where the first element is a string and the second element can be a string or an array of strings This allows
28. value L j 6 B Notes l The test identifiers for parameter t values are the numbers used to identify the parameters in the Values dialogs There must be a match between these identifiers and the parameter names appearing in the name field otherwise no EDF p value is returned the conventional t table is used This minimizes though does not eliminate the possibility of inadvertently using the wrong table for a test Identifiers for the diagnostic tests and test statistics are in most cases the numbers given in Tables 1 and 2 in the programming manual The exception is the case of user programmed tests which are given the numbers 1 2 etc This is with a view to backward compatibility since additional tests may be added to the list in future program releases Thus 1 is interpreted as the highest number on the list in Table 2 which in the current version of the program is 23 The interpretation of the test parameter depends on the case but in most cases it is the degrees of freedom of the test or number of restrictions If it is not used or has default value as in the case of t ratios it is assigned the value zero If non zero the value is appended to the statistic name in Row 1 enclosed in square brackets Different numbers of bins are permitted for different tests and different sample sizes This file format must be followed exactly or else the file cannot be loaded 37 James Davidson 2015
29. which can also be performed manually are to edit the executable file tsmod_run ox to include the following lines define TEXT INPUT import lt packages tsmod4 tsmgui4 gt Text Input include settings txt The compiler switch define INPUT instructs the program to read the settings After restarting TSM the file is edited again to remove the extra lines Note if performing this operation by hand e g in Linux that the f define statement must precede the import statement and the function should follow it similarly Note the run ID counter can be reset by including the line RUN ID value 31 James Davidson 2015 where the value is any nonnegative integer Use the special settings pull down menu in Options General to change the counter interactively To load two or more models in this manner where each has been saved to its own text file it will be necessary restart the program for each one The model listed in the file tsmod run ox will be read in as the current settings and can then be stored under the desired name in the Setup Model Manager dialog 32 James Davidson 2015 Appendix F Using TSM in the Classroom This appendix draws attention to the program features which may be of special use when using the program for teaching Please see the Users Manual for further details of all these features Installation and Use TSM licence information is stored in a f
30. 0 25 9 32 33 3519 41 07 33 85 3671 42 46 35 27 37 94 43 68 0 2 16 33 21 35 95 41 78 34 53 3749 43 27 36 07 38 77 44 09 0 15 32 11 33 9 3666 142 4 35 39 38 12 43 95 36 85 39 55 44 84 0 1 81 34 72 3748 43 1 36 19 39 05 44 52 37 64 40 38 45 89 0 05 361 35 81 38 53 44 2 37 24 40 05 45 39 38 68 41 36 46 69 45 James Davidson 2015 p 19 p 20 To 10 5 1 10 5 1 0 5 1 2719 30 14 36 21 2841 31 41 37 57 0 49 1 08 29 36 32 4 38 97 30 44 33 56 40 04 0 48 1 17 30 18 33 2 39 93 31 39 34 62 40 88 0 47 027 30 87 33 86 40 28 3201 3512 41 72 0 45 1 49 31 95 3482 41 4 33 00 36 41 42 97 0 4 2 25 33 54 36 53 43 34 34 78 37 93 44 2 0 5 3 45 34 71 3779 44 3 3604 3921 45 34 0 3 5 44 35 72 3877 44 76 37 04 40 25 46 14 0 25 9 36 6 39 75 45 66 37 92 40 98 46 79 0 2 16 37 47 40 43 46 68 3897 41 9 47 75 0 15 32 11 38 36 4125 47 25 39 96 43 48 4 0 1 81 39 16 42 01 47 95 40 74 43 76 49 58 0 05 361 40 27 43 05 49 02 41 83 44 52 50 76 p number of parameters Use the first column of the tables for the cases where the maximization interval bounds as proportions of sample size and 1 7 interpolating row values as required To approximate critical values for a test with lower bound and upper bound 7 compute _ 7 1 7 z l and find the nearest to this value in the second column of the tables interpolating row values as required
31. 2015 Program malfunctions If TSM should crash due to an Ox execution error it restarts automatically displaying a screen with instructions for saving diagnostic information The Ox error message including line numbers to identify the crash point is shown in the console window If the crash occurs in TSM code please copy the message and email it along with a note of the TSM version you are using to tsmail timeseriesmodelling com Please make sure you have the latest version of the program since bug reports on earlier versions cannot be considered The error screen displays options to continue or quit and also to export the current settings and models to a file called error_report tsm in the home directory It will greatly assist in diagnosing problems if this file is emailed along with the error message However please be aware in case there are any issues with confidentiality that this file will contain the data set and model specifications In rare circumstances the Java front end can freeze up with the program becoming unresponsive If this happens closing the DOS box running Ox usually terminates the program If the TSM window does not close at this point either close it in Task Manager or run the utility renewjava exe in the installation folder By default restarting the program should restore the complete working environment including the results window contents 9 James Davidson 2015 Appendix B Setting GUI Appearan
32. 3 23 63 19 56 2212 26 86 0 35 3 45 17 09 19 34 24 1 17 06 19 46 24 64 20 49 22 93 27 77 0 3 5 44 17 74 20 00 24 86 179 20 36 25 64 2127 23 65 28 5 0 25 9 18 38 20 63 25 11 18 61 20 95 26 1 21 93 24 31 2923 0 2 16 19 04 21 07 25 72 19 17 21 47 26 76 22 54 24 91 29 92 0 15 32 11 19 69 21 84 26 23 19 82 22 13 27 25 23 15 25 47 30 52 0 1 81 20 32 22 51 26 91 2045 22 87 27 69 23 77 26 16 31 15 0 05 361 21 00 2322 27 53 21 23 236 28 77 24 64 26 94 31 61 44 James Davidson 2015 p 10 11 12 To 10 5 1 1096 5 1 10 5 1 0 5 1 15 99 1831 23 21 1728 19 68 24 73 18 55 21 03 2622 0 49 1 08 1735 1979 24 62 18 65 21 02 26 43 2027 227 28 04 0 48 1 17 18 08 20 35 25 75 19 4 21 81 27 04 21 01 23 54 28 64 0 47 0 27 18 67 20 92 26 43 19 94 22 36 27 84 21 54 24 08 29 14 0 45 149 19 39 21 78 27 3 207 23 27 28 47 22 42 24 99 30 03 0 4 2 25 20 74 23 15 28 86 22 24 24 95 30 2 23 81 26 26 31 52 0 35 3 45 2187 24 17 29 76 2325 25 93 31 14 248 27 45 32 46 0 3 5 44 22 73 25 05 30 74 2419 26 71 31 65 25 63 28 18 3327 0 25 9 23 32 25 8 31 32 2482 27 43 32 39 26 37 28 98 33 77 0 2 16 24 2642 31 98 25 41 27 93 33 42 27 08 29 61 34 56 0 15 32 11 2462 27 00 32 33 26 14 28 55 33 75 27 74 30 16 35 07 0 1 81 25 39 27 87 32 95 26 86 29 21 34 46 28 47 30 88 35 75 0 05 361 26 24 28 63 33 86 2791 30 15 35 14 2931 3176 367 13 14 15 To
33. 5 Appendix C Including User coded Ox Functions There are several different ways in which Ox code written by the user can be incorporated into TSM See Section 4 6 7 of the main TSM document to see how these functions can be integrated into models formulated in the program e UserFunction return a vector matrix of equation residuals representing fi in quation 4 32 see the main TSM document Section 4 6 3 e UserSolve invert fi in equation 4 31 i e solve model from residuals generated by the program UserLikelihood return a likelihood function or other estimation criterion to be ptimized directly by passing TSM s built in options o e UserTest compute a test statistic from estimation outputs criterion gradient and hessian residuals parameters etc Two additional options bypass TSM s built in features e UserStatistic compute a statistic direct from the data set e UserGenerate generate a full sample of random observations The following function is to allow unspecified user modifications of various kinds e UserCode return a value depending a state of the program Basic Coding Guidelines With the main exception of User Test which has important extra arguments the basic format for these functions is similar For example the first case should take the form UserFunction const mcX const cStart const cEnd const vP const aName const bMode where the elli
34. 5 As with UserStatistic TSM program functions can be called from UserGenerate Controlling Program Functions In the present implementation the function UserCode const Input const Output can be used to control the acceptance of a Monte Carlo replication Out put represents a memory location allowing values to be passed by reference The line Output 0 x places the contents of x in a location where it can be accessed in the calling routine Currently this function is called after a Monte Carlo replication is complete Setting Output 0 1 causes the replication to be discarded similarly to a convergence failure in the optimization routine This allows criteria chosen by the user such as parameter estimates falling in an illegal region or other evidence of a false maximum of the estimation criterion to be used to decide whether to discard a replication To activate this feature go to Options General Special Settings and set Censor MC Replications to TRUE Note that the argument Input is not used in this application This function is designed so that it can to provide a similar functionality in other contexts in future releases 24 James Davidson 2015 Loading the Code To include the coded function or functions TSM should be started from the user s working directory The Windows installation sequence creates a special short cut TSM4 with User Code which runs the copy of the run file t smod_run ox
35. Running TSM In normal use TSM is started by running the Ox executable 0x1 exe on a starter file called t smod_run ox by default which in turn loads the main TSM code module This in its turn starts the Java Runtime Environment JRE There are always two icons on the task bar one with red TSM logo for the JRE the other representing the DOS box running Ox The latter normally runs in minimized mode but it displays error messages in case TSM terminates incorrectly There are three basic ways to initiate this sequence 1 By clicking a shortcut on the Start Menu or Desktop The installation program places two shortcuts on the Start Menu 4 James Davidson 2015 e Time Series Modelling 4 is for normal use and runs the Windows batch file tsmod runsc bat This starts the instance of t smod_run ox residing in the TSM Home directory TSM4 with User Code runs the batch file t smod_runuc bat which starts an instance of tsmod located in the Start in directory If this does not exist it will be created It can be edited to allow the user to compile his her own Ox code with the program The start menu icon associated with these shortcuts shows the TSM logo enclosed in a white square Since errors in the user s code can prevent TSM from starting it is advisable to keep both these start up icons available on the Start Menu 2 By double clicking a TSM settings file having t sm extension and red TSM icon in W
36. Time Series Modelling Version 4 47 Appendices James Davidson 23 September 2015 Appendix A Installatiois DAI E e c 3 3 Installations 3 64 bit TSM nenn 4 Starting and Running TSM iiss i Ferre teer aaa dase 4 About Home Folder d se etre eee vu ee aeu 5 Creating Additional Shortcuts and Start 6 Installing e neo dies erede a pai deriva bed edil v does 6 depend E A 7 Tiroubleshooil cs eios ress edocet Feu a s am reet 7 TSM under Linux and Mae OS X ae Sacco ret erede y itte ere una 7 SsfPack State Space Modelling a tin d bee bt eee ded fa Ponts 8 Program analfunctioDg su eque bon Poire pre aser a PUR MEHR et 9 Appendix B Setting GUI Appearance and Run Options 10 Appendix C Including User coded Ox Functions sssse 15 Basic Coding Guide lites aac tact ROREM 15 Bestdualso ILLO NE 17 Solving the Modele rrt tr terrena rb Ye oreet a e da A e c dee 18 Coding the Likelihood a eie etra erc oap PI iota iP dp a aait 19 Test
37. _Input PRINT_RESULTS 1 INPUT FILE data xls SERIES Garch INTERCEPT1 1 IS_ARFIMA 1 AR_ORDER 1 MA ORDER 1 IS GARCH 1 GARCH AR ORDER 1 GARCH MA ORDER 1 main Set Defaults Text Input Run Estimation TSM command variables upper case are globally defined and can appear anywhere in the program See the document TSM4 Programming Reference t smod4prg pdf for a full description Since TSM is a big program a command line switch is needed to reserve more memory than the default To run your program from OxEdit first do the following 1 Choose Tools Add Remove Modules 2 Select the entry amp Ox 3 Edit the Arguments field to read s6000 6000 S FilePath In other words add the s6000 6000 switch at the beginning of the entry 4 Close the dialog This setting will be remembered by the OxEdit installation This operating mode can be used in case the GUI installation does not work satisfactorily on certain systems It only requires a working Ox installation to run However its most important application is to allow TSM to be used as a programming module The estimation results can optionally be accessed within the user s program instead of being written to the console The program has commands to create batch jobs automatically for estimation runs and Monte Carlo experiments This facili
38. actual estimation sample see the main TSM document Section 13 3 for details EDF File Format It is possible to create EDF files by hand using data from different sources The format completely revised in Version 4 24 allows the EDF file to contain any number of columns in any order each representing the distribution or distributions given two or more sample sizes for a statistic The format of a column is as follows The first five rows contain identifying information for the statistic Row 1 Name for parameters and test statistics character string Row 2 Test type e 0 for value of a model parameter e 1 fora test statistic whole model including user coded statistics 2 for a diagnostic test statistic for equation residuals In multi equation models 2 denotes the diagnostics for Equation 1 In addition 3 for diagnostic test statistic for Equation 2 residuals e etc etc That is there are M 2 possible values for an M equation model Row 3 Test identifier see Notes Row 4 Test parameter degrees of freedom or other Row 5 S gt 1 the number of sample sizes for this statistic Then follow S blocks of rows The within block row values are as follows Row 1 R number of rows in this block 36 James Davidson 2015 Row 2 Sample size Row 3 0 reserved for future use Row 4 Minimum bin value L Row 5 Bin width B Row j for j 6 R 5 cumulative frequencies up to and including bin
39. alues etc However be careful to note that such procedures would not be valid in UserFunction UserSolve UserLikelihoodor UserTest 4 As for UserTest p values can be taken from the currently loaded EDF file if this option is selected see Critical Values Note 2 The name returned by the function must match one in the EDF file The natural application is of course to use an EDF file generated from a Monte Carlo simulation of the same test under the null hypothesis 5 Since the specifications in the Model Dynamic Equation dialog are bypassed by TSM in this case the dialog can be used to specify lists of variables which are passed to the code by external declaration see Coding the Likelihood Note 3 for details 6 It is possible to return comments which will be printed following the statistic values and p values Append up to m additional strings onto the array aName These can be of length 0 if there no comment These extra elements of aName are ignored when the function is called from a Monte Carlo experiment Generating Data The format for the function is UserGenerate const mcX const cStart const cEnd const vP const aName const bMode decl return mY The return value should be a column vector or matrix of generated data having cStart 4l rows 23 James Davidson 2015 Notes 1 The generated data are stored under the name s of the dependent variable s selected in the Model
40. atistics are all optional outputs although enabled by default See the Options Tests and Diagnostics dialog The minimal output consists of the regression coefficients sum of squares 87 and residual standard deviation Easy Equation Graphics Display The double chart button on the toolbar shows by default the time plots of actual and fitted values and residuals The option Extended Actual Fitted Residual Plots set in Options Graphics includes the actual fitted scatter plot and the histogram and kernel density of the residuals distribution These plots are also available as items on the Graphics menu but this option places them all together in a way the student can access easily Distributing Class Exercises The teacher can prepare a class exercise in the following way Read in a data set or generate an artificial one using the simulation module set up one or more model specifications and store these in Model Manager and also set other program options such as the display options described above Then give the command 33 James Davidson 2015 File Settings Export and assign an identifying name to the file This command creates a file with t sm extension in which all the data and settings are bundled It is also portable since it contains no local path information Local paths are stored by the command File Settings Save note Distribute the file to the class with instructions to copy it to
41. ce and Run Options Changes can be made to the appearance and behaviour of TSM by editing the file tsmgui4 oxh in the installation directory This file specifies certain external functions which return user selected values to the program Each function consists of a single line return option where option is the user selected item either a text string enclosed in or an integer value The main settings are made by running the installation program which edits this file automatically according to the user selections Editing the file manually allows fine tuning of the installation Edit these lines with care since errors may prevent TSM from starting Get HomeDirectory returns a text string containing the path to the TSM home directory which is the directory where your TSM installation resides By default this is the string returned by the Ox function ox ilename 1 but an explicit path can be substituted Use this setting if e g the help and graphics functions are not working as expected Get CWD returns a text string containing the path to the desired working directory By default this is the string returned by the Ox function get cwd An explicit path can be substituted to let files be written to a location different from the directory where TSM is started Get GnuPlotExe returns a text string the explicit path name of the installed Gnuplot executable If this points to a separate Gnuplot installation it w
42. d as described in Programming Reference Section 5 Note that the latter items are not sorted alphabetically but are listed in the order given in the Programming Reference Also note that stored models are not saved To save a stored model in text form first load it in the Setup Model Manager dialog then save it to an identified text file using the command File Settings Settings as Text Save Current Model Use a name such as Model txt where is the model identifier To use these exported settings in an Ox program calling the TSM kernel as described in Appendix D simply copy the contents of the file to the clipboard and paste the lines into the ox file between the braces ofthe Text Input function The order of the statements is arbitrary However an even simpler method is to create the function Text Input include settings txt The include statement inserts the contents of the named text file into the program at run time The path to the file should also appear if the file is not located in the working directory The settings contained in the Text Input function can also be loaded into the GUI so that a prepared batch job can be run and modified interactively To do this all that is necessary in Windows is to create a file such as sett ings txt as above give the TSM command File Restart Load Text Input and select the prepared file in the file dialog The actions carried out by this command
43. de and documentation files to the TSM home directory It is easy to modify the TSM start up script batch file to check for updated files on a central server and download these as required as part of the start up sequence For example after setting the required write permissions edit the file tsmod_runsc bat and add lines at the top similar to the following net use y server tsmod xcopy q d s y z y c Progra 1 OxMetrics5 0Ox packages tsmod4 Here server is the name of the central server and tsmod should be replaced by the path to the TSM file store The assigned drive letter should also be set as required Upgrading a network installation is then a simple matter of copying the new files to the central server either from the zip download or from an installation that has been updated by the tsm setup exe download 34 James Davidson 2015 Appendix G Using Empirical Distribution Functions TSM can both create and use distributions in the form of cumulative frequency tables for calculating test critical values and p values These tables are stored in spreadsheet files If tables for the same statistic in different sample sizes are created they will be interpolated to give an approximate p value appropriate to the actual sample size see the main TSM document Section 13 3 for details of the method Creating EDFs The usual method of creating an EDF is by a Monte Carlo experiment to simulate the
44. distributions of statistics under the null hypothesis There are two ways to have TSM make an EDF file from Monte Carlo data Checking the box Save EDFs in the Setup Monte Carlo Experiment dialog will result in the raw cumulative frequency tables being written after each run with names of the form EDF Runfi ext Alternatively a more flexible approach is to open the Graphics Monte Carlo Distributions dialog after the run is completed Pressing the Make EDF File button opens the file dialog to save the EDF file which will contain tables for all the statistics simulated in the experiment The empirical distributions can be plotted in this dialog as frequency histograms with a superimposed kernel estimate of the density Adjust the kernel bandwidth using the scrollbar re displaying the plot to see the result It is the cumulated forms of these kernel estimates which are used to create the EDF so check and adjust as necessary all the tabulations before creating the file EDF files created for the same statistics with two or more sample sizes can be merged into a single file A file on disc may be merged with the table in memory using the command Files Data Tabulations Merge EDF File The resulting combined file is optionally saved under a name supplied by the user otherwise under the name of the file in memory Notes 1 Remember that Monte Carlo distributions are stored along with the other results under the data
45. e which opens the file dialog Suppose the file mycode ox is selected Then the following actions take place 1 usercode ox is edited to contain the line include mycode ox Any other include directives are deleted and any other contents are commented out 2 TSM is restarted loading the new code Since usercode ox is a system file whose name is cannot easily be changed this strategy makes it much easier to load different bits of code for different purposes usercode ox can also be edited by hand Linux users do not have the Restart feature in the current version If the code file is located in a different directory include the path as well as the file name in the quotes but remember that Ox path formatting conventions must be used avoid single backslashes Windows can handle either V or the latter is also valid under Linux Note Previous versions of TSM used separate define directives for each type of function for example In Ox a include compiler directive followed by the path and name of a text file in double quotes has the same effect as if the contents of the named file were inserted at that position 25 James Davidson 2015 define USER_FUNCTION This type of scheme can still be implemented by editing t smkn14 h In this case the dummy do nothing functions can be deleted from the code file However it is then more difficult to implement a function library and export and i
46. e Code at Start up or created if it does not exist 4 TSM program is restarted to compile the code Notes 1 Ifa file of the same name as the loaded file already exists in the Start in directory the name of the new file is changed as necessary to be unique For example if mycode ox exists the new include file is named mycodel ox and the A line of code is commented out by placing the characters in front of it Delete these to uncomment 27 James Davidson 2015 include directive points to this If mycode1 ox also exists the new file is named mycodel11 ox And so on 2 If tsmod run ox exists in the Start in directory it is edited automatically to include the required compiler directives Any pre existing define statements or other lines or comments added by the user to should be preserved at Step 3 above but check this file in a text editor in case of unexpected behaviour 3 It is possible to have the exported code contained in usercode ox itself although this strategy is not recommended In this case its existing contents are commented out before the new code is added Documenting the Code TSM features a menu command Help View Files Imported Ox Code which displays the contents of the current code file either usercode ox or the first file included in it in a similar format to the Help pages The natural way to document functions is therefore to include the explanatory text at the top of
47. e jobs currently queued on the system append the argument g to the command condor_rm This command terminates running Condor jobs Append the cluster number as an argument to the command to remove a particular cluster or a11 to clear all your jobs condor release Releases jobs in Held mode To remove such jobs from the system run this command first appending the cluster number or all condor submit The command to launch a job or a set of parallel jobs It must be followed on the line by the name of the submit description file which by default in TSM is condor_submit txt Normally there is no need to give the condor submit command by hand since TSM performs this step automatically after creating the submit description file However jobs can be deferred to be run manually and of course Condor can also be used independently of TSM to run other programs To check out the syntax of the submit description files inspect condor_submit txt as created by TSM A copy can be found in the working directory following a run Special Settings The Special Settings pull down menu in the Options General dialog contains four interactive settings that are visible when Condor is enabled These specify commands appearing in the condor_submit txt file They are Condor Notification Sets the conditions for notification of job status by email Condor Email Email address for notification Condor Output Files If TRUE the con
48. e working directory or the TSM installation directory This file if it exists will be run automatically prior to launch A possible requirement under item 4 is to wake up a cluster of machines from power saving mode Consult your network administrator about these requirements How Condor Works Various configurations of the Condor system exist for different roles This paragraph explains how Condor is implemented from the standpoint of TSM There are three steps l All the files needed to run an Ox job including Ox executables the Ox source file for the job and the data file are copied from the local machine to the Condor server TSM automates the launch procedure by creating a submit description file called condor submit txt which lists the required files and their locations and contains other instructions for the run such as the executable to be run the argument list and so on The Condor server finds a workstation on the network to run the job sets it up there and launches it Periodic snapshots of the machine s state are recorded If another user interrupts the run by pressing a key or clicking the mouse at the target workstation Condor suspends the job moves it to another free workstation and resumes execution at the point of the last recorded snapshot The output files written by Ox are written to your working directory appearing exactly as if you had run the job locally Optionally an email is sent to the us
49. er to notify completion of the job For Monte Carlo jobs in parallel these outputs are special data files with the t sd extension When all the parallel instances have returned their outputs the Results button in the TSM Monte Carlo dialog can be used to load these files aggregate the results and display them as the output of a single experiment View this file after submitting a job to see how installation settings 2 and 3 above are passed to Condor as assignments to the Boolean requirements variable 39 James Davidson 2015 Command Line Interactions To interact directly with Condor open a DOS box This can be done by choosing Command Prompt from the Windows Accessories menu or by clicking Start then Run and entering CMD in the Open field The following are the basic commands to control and get information from Condor Type them on the command line and press Enter For more information on these and the other available commands see the Condor user s manual downloadable from http www cs wisc edu condor condor status Displays list of the machines available to take Condor jobs and their status Machines don t appear on the list if they are switched off or in power saving mode condor_q Shows the status of the currently submitted jobs and their status When a job is terminated it disappears from the list By default the command shows just the jobs run under your username To see all th
50. file messages txt in the Start in directory Any other suitable file name can be substituted This would be an ideal scheme to implement routinely but unfortunately Windows SP1 and SP2 contains a bug see Microsoft Knowledge Base Article 886659 that can cause an unpredictable lock up of the redirection file TSM then cannot be started without either rebooting or designating a different redirection file The only way to see if your system is affected by this problem is to try it However since the 28 James Davidson 2015 problem appears not to arise immediately this is probably the quickest way to get specific diagnostic information Remember to remove the redirection switch if it causes a problem 2 Running in a DOS Console a Use text editor to create a batch file in the Start in directory called say start tsm bat b Enter the lines cd start in ox home packages tsmod4 tsmod_runsc bat ox home Nbin start in Here ox home and start in stand for the Ox home directory and Start in directory paths as entered during the installation sequence c Opena DOS box Start Run Cmd or go to Command Prompt in the Programs Accessories menu d Atthe prompt enter the commands Start in Xstart tsm bat This starts TSM normally and any Ox messages will appear in the console window Note simply executing the batch file by double clicking it in Windows Explorer also launches the program but in this ca
51. generation model for the experiment To get access to the required tables load this model in Model Manager Don t use the model for further experiments 1 save it under a different name before re using the original if you want to keep the associated tabulations which are stored in the associated t sd file 2 It s a recommended practice to prefix the names of EDF files with to distinguish them from other spreadsheets 3 The files created as described contain EDFs for all the eligible tabulations that is t values and test statistics although not the estimators in non normalized form If any of these tables are not required the columns should be manually deleted from the file either using a spreadsheet program or by loading the file into TSM as data and using the Save Selected command in Setup Data Transformation and Editing Edit 4 Setting the kernel bandwidth to 0 yields the raw table in other words the cumulated histogram This is also the form of the distribution that is saved by checking the Save EDFs box in the Monte Carlo dialog There are no established results to indicate what degree of smoothing gives the best approximation to the tail areas of the true distribution This must be a matter for the user s judgement There is no 35 James Davidson 2015 substitute for increasing the number of Monte Carlo replications to estimate a distribution more accurately 5 It is not possible to cha
52. h estimation These warnings are not under the control of TSM and appear in the Ox console window not the TSM results window Note that echoing of TSM output to the console is controlled by an option in the Options General Options dialog Here is the list of key words identifying lines that will be preserved when the file tsmgui4 oxh is updated Get HomeDirectory Get CWD Get ResultsBufferSize Get HelpDir Get IconDir Get NameCharacter Get LFType 13 James Davidson 2015 Get_FontSize Get_ChFontSize Get_FontStyle Get_TextSize Get_TextStyle GraphDelay Get_LenMRUList Get_BgColor Get_HgColor Get_TextAreaBG Get_LenMRUList Get_LenMRUModList Get_MaxDatsets Get_StartFileName Get_CodeFileName Get_BatchFileName B Get_SettingsFileName Get_ModelFolder Get_EDFFolder Get_BatchFolder CondorNetOS CondorNetArch Get ComboBoxRows Get GnuPlotExe Get 64Bits Get IconSize WidthFactor ChangeUserPath define GNUADJUST define TEXT INPUT define TEXT DEFAULTS define OXDRAW define OXWARNING define CONDOR define USER_CODE define SSFPACK Any line containing one of these strings is passed through unchanged to the updated file when the installation is upgraded Additional settings for use when the program is run as an Ox module not in GUI mode should appear in tsmkn14 oxh 14 James Davidson 201
53. he speed of execution and is recommended as the default setting when writing Ox functions Omitting the qualifier has the effect that a local copy of the argument is created with that name and this can be changed although the argument itself existing outside the function is not changed 2 Observe the Ox convention that matrix row and column indices start at zero If the whole sample has been specified for the calculations then cStart 0 and cEnd rows mcX 1 A matrix is indexed with row and column indices appearing in separate pairs so that the tj element of the data matrix is mcX t j 3 VarNum isa TSM function that converts a variable name a text string enclosed in double quotes to the corresponding column number of the data matrix Note that variable names are case sensitive 4 Variables in the data set can be referenced directly by name using the VarNum function or alternatively program model settings can be referenced using the variable names defined in the TSM Programming Manual For example the dependent variable s selected in the Dynamic Equation dialog can be accessed in SERIES which is an array of one or more strings names The column of the data matrix containing the first or only dependent variable is obtained as VarNum SERIES 0 5 Functions can call other functions which can call yet others The rules of Ox programming apply but the simplest rule to observe is that a called function must
54. he home folder location itself is not stored in the settings file so if a different settings file is opened while the program is running the home folder is unchanged However if TSM is started by double clicking a settings t sm file in Windows Explorer the home folder is the location of this file at start up Other file storage locations are stored as program settings so the program will continue to look for its data in the places remembered from the last session regardless of where it is started However these locations can be changed as required using the menu commands Files Folders The files other than results files that are not accessed by the program and batch files which are temporary are moved automatically to their new locations If a settings tsm file has been created using the command Files Settings Export all relevant material data model files tabulations etc is bundled in the file and existing storage locations are not preserved unless they are subfolders of the home folder Thus such files can be freely ported between installations and distributed to other users At start up all stored files are written to the same locations relative to the home 5 James Davidson 2015 folder The option to move one or more types of file to new locations can then be chosen as before Creating Additional Shortcuts and Start Options Here s how to create a new Windows shortcut from scratch a Open the
55. ibution with iDF1 1 to allow for trend Dist 9 Lo s R S distribution Dist 10 Nyblom Hansen distribution for iDF 1 parameters Dist 11 Kiefer Vogelsang Bunzel F distribution for iDF1 parameters To return gt 2 tests simply concatenate the columns for each test and return a 4 matrix The additional function arguments are as follows where T cEnd cStart 1 and m number of equations vParam The full parameter vector 1 p r mGradients Gradient contributions T x p mHessian Hessian matrix p mCovmat Covariance Matrix p p formula as specified in Options Tests and Diagnostics vRes Residuals T x m vSigmas Conditional variances T x m CV models only aDstat Diagnostic statistics an array of dimension m of 1 x 13 vectors mStat Specification tests 1 x 20 Notes 1 vParam contains all the parameters in the model including p estimated elements and r fixed and solved elements The position of a parameter in the vector is found as the number in the left hand column in the relevant Values dialog vP in the functions defined previously includes only those appearing in the user defined function hence is a subvector of vParam The column of mGradients or mHessian corresponding to an element of vParam has to be found by subtracting the number of restricted fixed or solved elements preceding it in the list from its number in the Values dialog The elements
56. ifferent from that on the machine running TSM this information must be provided to the system In tsmgui4 h edit the line CondorNetOS return so that the returned string identifies the operating system The codes are listed in Appendix A of the Condor manual under OpSys in the section headed Machine ClassAd Attributes For example in Condor 7 8 1 the string is WINDOWS for any version of Windows The returned string can be replaced by an array of strings if 38 James Davidson 2015 more than one operating system is represented in the cluster older versions of Condor needed to distinguish XP Vista W7 and so forth If the system architecture of the machines of the Condor cluster is different from that on the machine running TSM this information must also be provided In tsmgui4 h edit the line CondorNetArch return The codes are listed in Appendix A of the Condor manual under Arch in the section headed Machine ClassAd Attributes In Condor 7 8 1 the supported options are the default INTEL and X86 64 Typically this setting will need to be changed from the default when the user s machine has 64 bit architecture and the machines of the Condor cluster are 32 bit Depending on the installation it may be necessary to run one or more programs as a preliminary to launching a Condor job Enter the required commands in a Windows batch file called condor_init bat located in either th
57. ile called registration txt in the TSM installation directory or home folder for users who do not have write permissions there To licence a copy on a workstation an alternative to typing in the username and key at each installation is to copy this file once created directly into the TSM directories of each installation This is the natural procedure when installing centrally over a network If a licence file registration txt is present in the user s home folder work folder it will take priority over the installed licence and this name will be displayed at start up This makes it easy for a user to run the program using their personal licence in case all the copies covered by the site licence are in use Linear Regression Mode Uncheck Options General Enable Optimization Estimators and then restart the program This option suspends the nonlinear optimization features that are used chiefly by more advanced practitioners The Running Man and Space Shuttle buttons are hidden and unnecessary dialogs disabled This turns TSM into a simple regression package and hopefully makes navigating the interface easier and less confusing for beginners Only numerical optimization features are disabled which includes all maximum likelihood options The bootstrap and Monte Carlo options are all still available Simplified Output The log likelihood and information criteria residual higher moments and Jarque Bera statistic and Q st
58. ill be used in place of the compact supplied version Note that this function is also defined in tsmkn14 oxh It should be edited separately if TSM is to be run without the GUI Get ResultsBufferSize returns the desired memory bytes reserved to contain the contents of the results window If the buffer is filled the user is prompted to save the contents to a file before clearing it The default value is 10000000 10 Mb This should be large enough for most sessions but raise it if you have plenty of memory Get HelpDir and Get_IconDir return the locations of help and icon files as sub directories of the TSM home directory Get_LFType selects the look and feel of the GUI Here are the options available for Windows e lor2 Windows look and feel depends on Windows version e 3 Windows Classic look and feel Oor4 Java Metal motif Note this setting should normally be changed by running the installation program This is the recommended procedure You will need to reboot your computer or close javaw exe in Task Manager after changing this setting 10 James Davidson 2015 Get IconSize selects the size of the toolbar icons Returning 0 selects small 16 x 16 pixel icons while returning 1 selects 32 x 32 pixel icons WidthFactor controls the on screen dimensions of all program objects including the frame dialogs and fonts When the installation executable is run setting Icon and Font
59. indows Explorer This action runs the associated batch file Start TSM bat which in turn runs the batch file tsmod_runsf bat passing the name of the chosen settings file In this case the Start in directory is the one where the selected tsm file resides The second icon on the task bar representing Ox is in this case the Windows black and white c prompt icon 3 By double clicking TSM model listings file having t sd extension and blue TSM icon in Windows Explorer Similar to 2 except that the file settings tsm in the current Start in directory is loaded if it exists and then the model or listing data are loaded About the Home Folder The home folder or work folder also referred to here by its Windows name the Start in directory is the storage location of the TSM settings file default name settings tsm from which program data are read at start up and to which they are written at close down and periodically during program operations By default other files generated by the program particularly data files associated with models and batch operations with extension sd are also stored in the home folder although other locations can be selected when the program is running In normal operations when the program is started from a desktop icon or Start Menu icon the home folder is the one selected during the installation procedure the path being used to create a Windows shortcut See the next section for details T
60. ion as required by fixing the corresponding parameters at 0 interactively in the Values Equation dialog This allows the specification to be changed and tested without needing to stop the run and re code 17 James Davidson 2015 Solving the Model If ex ante forecasts or stochastic simulations are required it is necessary to code the inverse of UserFunction to retrieve the original series back from the residuals The function UserSolve must be created to do this This function must take the form UserSolve cons const vP const mcX const vcY const cStart const cEnd aName const bMode Note that it takes one additional argument which is the vector or matrix for systems of residuals Here is the bilinear example UserSolve const mcX const vcY const cStart const cEnd const vP const Notes aName const bMode decl vcx zeros cEnd cStart 1 1 decl xnum VarNum Bilin for decl t cStart t lt cEnd ttt if t gt 0 if cStart gt 0 vcx t 0 vcY t 0 vP 0 vP 1 mcX t 1 xnum vP 2 mcX t 1 xnum vcY t 1 0 return vcx This function should return a vector of values or a matrix for systems with cEnd cStart 1 rows For simulations it is called with cStart cEnd to provide one new value at a time cStart 0 is a possible value The coding must be designed to allow for this returning e g zero if lags are not available bMode
61. ity or log probability terms The function maximized is the sum of the terms returned The program may also differentiate this vector term by term by difference approximation to compute score contributions Notes 1 It is possible to compute residuals simulations and forecasts in this mode provided UserSolve and UserFunction are defined in the usual way If UserLikelihood itself calls UserFunct ion it only needs to supply the code to transform residuals into log likelihood contributions 2 UserLikelihood can return a name similarly to UserFunction This is ignored by TSM unless bMode 0 3 Because the built in model features are bypassed in this case the Dynamic Equation dialog can be used to specify variable groups which can then be accessed by the code see Note 4 under Basic Coding Guidelines In addition to SERIES the arrays of names that can be accessed include REGRESSORS_1 REGRESSORS_2 REGRESSORS_3 and INSTRUMENTS This can allow the model specification to be changed on the fly without changing the code Coding a Test The UserTest function receives as arguments the main outputs of a successful estimation run These can be used to code one or more test statistics to be written to the Results window if this option is selected in Model Coded Function To return a single test the format is UserTest const vParam const mGradients const mHessian const mCovmat const vRes const vSigmas aDstat mStat const
62. m http www gnuplot info A compact Windows version of GnuPlot is bundled with TSM but installing the latest release is recommended It will be used if it is present A graphics file viewer processor is a useful accessory 4 Under Windows IrfanView from http www irfanview com is recommended freeware IrfanView will display png gif eps and emf files and allows limited editing and conversion to other formats such as jpg Also ideal for your holiday snaps A PDF viewer is required to read the documentation files If not already installed on your system get 5 Acrobat Reader freeware from http www adobe com products acrobat readstep2 html For data input suitably formatted text files can be used at a pinch but a spreadsheet format such as x1s wks or in7 GiveWin format is recommended 15 Microsoft Excel format is the TSM default data format Regrettably a Linux installation is no longer available since we do not yet have a Linux compilation of OxJapi Version 2 compatible with Ox 7 Volunteers with Linux or OS X expertise and tools are eagerly sought Meanwhile TSM for Windows can be run under Linux or OS X using a package such as Wine see http www winehq org download Installation This is by running the executable setup file Be sure to run the TSM installation after installing Ox and the JRE The TSM files are installed in a sub directory of the Ox installation which must accordingly exist The
63. ments decl ia ib str sscan amp aName 0 s amp str i amp ia s amp str i amp ib Note that the format s reads everything up to the next space character as a string Spaces can therefore be conveniently used to separate the name components The name is entered in the GUI by typing it into the Ox Test Name field or selecting from a list see Maintaining a Function Library below in the Model Coded Function dialog see the User s Manual for details Observe that the text components are arbitrary provided the order of components separated by spaces is maintained Critical Values The program prints p values for tests based on known formulae in cases in cases 0 5 of the distribution codes Cases 6 10 use published tabulations of critical values obtained by simulation for specified tail probabilities In these cases since the tabulations are only for selected significance levels the p values are reported in the form of inequalities It is also possible for the user to supply critical values for a test either from published sources or generated by Monte Carlo simulation in TSM There are two ways to do this 1 Return a set of critical values with the statistic In this case the function User Test should return a matrix with seven rows and one column for each statistic Thus in the case of a single test the return statement might appear as follows return dStat 1 dCV50 dacV10 dCV5 dCV2 dCV1 where the elemen
64. mport code Adopting the scheme described here is strongly recommended Maintaining a Function Library By preparing the code file suitably it is possible to run a coded estimator or test by simply selecting the function from a list in the Models Coded Function Test dialog To implement this scheme an additional function to return a list of names must be created for each function type The naming functions are UserFunction Names UserLikelihood Names UserTest Names UserStatistic Names UserGenerate Names These functions take no argument and should return an array of strings These names are displayed in the Models Coded Function Test dialog when the corresponding Ox function type is selected and can be displayed in turn with the Previous and Next buttons The displayed name more precisely a pointer to its location in memory is passed to the function through the aName argument The function must accordingly be set up to return the required output depending on the value of this string For example suppose two test statistics have been programmed to be called MyTest1 and MyTest2 In this case the naming function should take the form UserStatistic Names return MyTest1 MyTest2 The function UserStatistic might now take the following form UserStatistic const vParam const mcX const cStart const cEnd const aName const bMode decl names
65. nces of a Monte Carlo experiment are run for subsequent aggregation the results are returned as a group of files in special format with the t sd suffix which the user must load into the program for aggregation and display A note about Condor terminology A group of jobs that is launched as a single submission to Condor is known to Condor as a cluster The individual jobs in a cluster are known to Condor as processes This is slightly confusing since the term cluster is also commonly used to describe a group of linked computers on a network In the case of TSM we refer to the instances of a TSM Every Condor job is identified by a cluster number but please note that this is different from the Run ID number assigned by TSM However the Condor process numbers do match the TSM instance numbers which run from 0 to N 1 when there are N instances This appendix gives details about running Condor on a Windows network on which we have gained experience Condor is implemented under other operating systems and could certainly be used in conjunction with TSM under Linux but we cannot give specific advice about this at this time Later hopefully Installing TSM for Condor To enable Condor in a TSM installation some changes must be made to the header file tsmgui4 h as follows 1 Intsmgui4 h uncomment the line define CONDOR 2 Ifthe operating system variant installed on the machines of the Condor cluster is d
66. nge the smoothing of a distribution once it is saved in EDF format Hence if in doubt retain the original tables as described in 1 6 Results for different tests also be combined using the merge command This happens automatically if the test names and identifiers do not match up across the files Using EDFs for inference Load an EDF file into memory with the command File Tabulations Load EDF File The loaded file can be used to calculate individual p values and critical values and also to create density plots in the Setup Look Up Tail Probability and Setup Look Up Critical Value dialogs To use the tabulations for model inference check the box Use EDFs from File for p Values in the Options Tests and Diagnostics dialog Then if an EDF file is loaded and also contains a name and identifier matching the model to be estimated the tabulation is used to calculate p values Otherwise the conventional tabulation is used It is ultimately the user s responsibility to be sure that the correct table is used for each test The EDF data resident in memory when a model is stored in Model Manager are stored in the associated t sd file with the other model information and reloaded when the model is loaded The original file does not need to be present once the data are associated with a model in this way If more than one sample size is tabulated for a statistic the tables are interpolated to give the best approximation for the
67. options presented by the program to specify the installation are as follows 1 The installation directory TSM Home This must specify a path of the form ox home packages Nt smod4 where ox home is the location of the Ox installation and in the most usual case will stand for c program files OxMetrics7 Ox Ina 64 bit Windows 7 8 10 system for 32 bit Ox including Ox Console the destination directory is c program files x86 OxMetrics7 Ox 3 James Davidson 2015 2 The Start in directory or home folder which will be used by default to write program outputs This should normally be a subdirectory of the user s Documents directory The user must have write permissions in this directory although once TSM is installed write permissions in TSM Home are not required 3 The Ox installation options are 32 bit Ox Professional or Ox Console and 64 bit Ox Professional Be sure that the selection at Stage 1 matches the selection made here Note that Ox Console the free version for academics is currently only available as a 32 bit application 4 The Start Menu folder by default Time Series Modelling 4 Choice of look and feel for the graphical user interface The Windows Classic interface is similar to that used in TSM versions up to 4 26 featuring 3D effect buttons and boxes Windows Standard is said to adapt itself to the flavour of Windows installed XP Vista etc Java Metal is described
68. psis represents the user s code Note that Ox is case sensitive The arguments passed to the function are as follows e is the data matrix with the series arranged by columns cStart and cEnd are first and last of the block of rows of mcX for which the calculations are to be performed e vP is the row vector of parameters as named in the dialog Model Coded Function e aName is a pointer to a location containing the string entered in the Function Name field in Model Coded Function The location itself a string on entry but can be an array of strings on exit is accessed as aName 0 e bMode is a Boolean flag set equal to 1 if the function call is being made on a second or subsequent occasion in an estimation or simulation run and 0 otherwise Consider the following example which generates the residuals for a first order bilinear model UserFunction const mcX const cStart const cEnd 15 James Davidson 2015 const vP const aName const bMode decl xnum VarNum Bilin decl zeros rows mcX 1 for decl t cStart 1 t lt cEnd vcy t mcX t xnum vP 0 vP 1 mcX t 1 xnum vP 2 mcX t 1 xnum vcy t 1 l if cStart cEnd return vcy cStart cEnd else return 0 Notes 1 The const argument qualifier means that the argument cannot be changed by the function Attempting to assign a new value produces an error This improves t
69. rly to INPUT this directs that default settings to be re instated when the File Settings Clear All command is given in the program are to be read from an external function called Text Defaults DJUST This allows some additional control over plot styles In addition to setting the define GNUADJUST directive add the following lines in the run file Adjust Plotsettings PLOTLINE 1 De first variable PLOTLINE_2 7 second variable PLOTLINE 3 5 forecasts PLOTLINE 4 3 forecast s e bands The settings shown here are the defaults For the alternative settings see the Ox or GnuDraw help documents Table drawl shows the available line styles and the DrawZ function shows the s e band styles Note line options for series plots are set as graphics options in the program Use this option to change the line style for graphics such as correlograms spectra QQ plots etc OXDRAW Loads the OxDraw drawing functions instead of GnuDraw With this option interactive display of graphics using Gnuplot is not available but gwg files can be saved to disk These can be displayed in a GiveWin or OxMetrics window and modified using the GiveWin OxMetrics graphics editing features OXWARNING Switches on Ox error messages that are normally suppressed such as warnings of attempts to invert singular matrices failure in eigenvalue routines etc These may be useful to diagnose problems wit
70. se the console window shuts on exit so cannot relay error messages 3 Running under OxEdit a Install Ox as an OxEdit module if this has not already been done Go to View Preferences Add Predefined Modules and choose Ox b Choose View Preferences Add Remove Modules and highlight amp Ox in the list Edit the Arguments field to read s6000 6000 S FilePath In the Initial Folder field enter the path to the Start in directory Close the dialog a Ph Load the file t smod_run ox from the Start in directory and launch it by choosing Modules Ox Messages from Ox including compilation and execution error messages are printed in the OxEdit window Note on Give Win OxMetrics In principle TSM can also be run from GiveWin OxMetrics using OxRun but this mode of operation is not recommended because the GUIs of each program do not co exist very happily However GiveWin OxMetrics is an excellent tool for organizing your data and its in7 format can be read by TSM It can also be used for graphics processing The OXDRAW compiler directive allows TSM to write gwg files See Appendix B for details 29 James Davidson 2015 Appendix D Calling the Code from an Ox Program TSM can be operated by a file of text commands without loading the GUI Most features including graphics are currently available in this mode The following shows a typical run file import packages tsmod4 tsmknl4 Text
71. sers subdirectories Using Q4Wine download and install the Windows versions of Java Ox console and then TSM and also Gnuplot by running the installation executables in the usual way in Q4Wine Explorer There does not appear to be a facility in Wine to create shortcuts to run batch files nor can TSM be started by the usual Windows procedure of double clicking a settings file Hence a different strategy must be used to start TSM Running ox1 exe in the Q4Wine command window is a possible method but a much more convenient alternative is to use OxEdit The one additional installation step needed is to use Q4Wine Explorer to access the TSM home directory and copy the file t smod_run ox into the working directory designated when TSM was installed The steps to start the package are then as follows 1 Navigate to the working directory in Q4Wine Explorer and double click on tsmod_run ox into start OxEdit with this file loaded 2 Runtsmod_run ox in OxEdit to start TSM Once TSM is running it can be operated in the usual way Either the operations of dragging and dropping from the Q4Wine Explorer or the usual menu commands can be used to load data and settings files Speed penalties appear negligible There are two further features of the true Windows implementation of the program that Wine in the version under examination cannot emulate The first a fairly minor issue is that the Courier typewriter font appears to be unavailable to Java
72. tents of standard output are returned in a file Condor Log Files If TRUE a log file is returned 40 James Davidson 2015 Appendix I Tables for Nonstandard Tests I QUANTILES OF THE DICKEY FULLER DISTRIBUTION prob of a smaller value The three cases raw data mean fitted mean and trend fitted lt 01 025 050 100 90 95 975 99 T 2 58 2 23 1 95 1 62 0 89 128 1 62 2 00 Th 3 43 312 2 86 2 57 0 44 0 07 023 0 60 Tr 3 96 3 66 3 41 3 12 1 25 0 94 0 66 0 33 ASYMPTOTIC CRITICAL VALUES FOR ADF AND PHILLIPS PERRON COINTEGRATION TESTS MacKinnon 1991 Table 1 See original table for small sample corrections Intercept Intercept and Trend P t lt v 10 05 01 10 05 01 No Regressors 2 5671 2 8621 3 4335 3 1279 3 4126 3 9638 1 Regressor 3 0462 3 3377 3 9001 3 4959 3 7809 4 3266 2 Regressors 3 4518 3 7429 4 2981 3 8344 4 1193 4 6676 3 Regressors 3 8110 4 1000 4 6493 4 1474 4 4294 4 9695 4 Regressors 4 1327 4 4185 4 9587 4 4345 4 7154 5 2497 5 Regressors 4 4242 4 7048 5 2400 4 6999 4 9767 5 5127 IV RS TEST QUANTILES OF THE DISTRIBUTION F y v A W Lo 1991 Table IT P V v 005 025 050 100 200 300 400 500 v 0 721 0 809 0 861 0 927 1 018 1 090 1 157 1 223 P V lt v 543 600 700 800 900 950 IS 995 v 1294 1 374 1473 1620 1747 1 862 2 098
73. ter returns a string containing the separator to be used for determining where the name of a variable ends and the appended description text begins Use this character or character string when preparing data for input in a spreadsheet or text file The default setting is 8 Get LenMRUList returns an integer the maximum length of the Most Recently Used file lists The default setting is 10 Setting to 0 turns off the MRU list feature Get MaxDatsets returns an integer the maximum number of data sets that can be stored in memory The default setting is 10 Setting to 0 turns off the option to store additional data sets Get LenMRUModList returns an integer the maximum length of the Most Recently Used model lists see Model Load a Model The default setting is 20 Setting to 0 turns off the quick model loading feature Get ComboBoxRows returns an integer the number of items to be displayed in a combo box pull down menu choice widget The default setting is 20 Get StartFileName returns the name of the file that is run under Ox to start TSM TSM needs to know this for restart operations The default name is 11 James Davidson 2015 tsmod run ox It be changed but the name must also match that appearing in Windows batch file scripts otherwise TSM will not start Get CodeFileName returns the name of the file used to contain the user s Ox code for compilation under TSM The default name is
74. ts are the 50 10 5 2 5 and 1 critical values If not all of these are known then replace them by the next largest value available For example if only the 10 5 and 1 values are available the return statement should read return dStat 1 dCV10 dCV10 dCV5 dCV1 daCV1 21 James Davidson 2015 This setup ensures that p values are still correctly reported in the form of inequalities 2 Supply the program with a spreadsheet file containing a complete empirical distribution function EDF This can be constructed in any desired way provided the format is correct but the TSM Monte Carlo module can create EDF files with the right format from a simulation of the null hypothesis See Appendix G for details of the file format Load the required tabulation file File Data Load EDF and check Use EDFs from File for p Values in Options Tests and Diagnostics The program must find a statistic name in the EDF file to match the statistic name returned in aName 0 This condition will of course be fulfilled if the tabulation is prepared using the same Ox code in a Monte Carlo simulation of the null hypothesis In case there is no match the program checks for distribution details as in Case 1 and uses these details if present Otherwise no p value is reported Note there is no need for UserTest to actually compute a test statistic It can simply retrieve one of TSM s built in tests from the arguments aDst at or mStat and return
75. ty allows large jobs to be run concurrently while keeping the GUI free for other tasks Studying the Ox files created by these commands may also be helpful in learning how to create more elaborate programs 30 James Davidson 2015 Appendix E Saving and Loading Batch Settings in the GUI The program settings in GUI mode are saved by default into a file with extension tsm but this file is only readable by the program To save the current settings except defaults into a text file in a format suitable for creating the equivalent batch job as in Appendix D use the TSM command File Settings Display Save Text This command allows creation of a text file whose contents has the format of the function Text_Input As well as providing a quick way to set up a batch job this is much the easiest way to learn to program using the TSM scripting language Set up a job interactively in the GUI and then write out the batch settings that correspond to it An alphabetical index of TSM commands can be found in the Programming Reference The file created by File Settings Settings as Text Display Save Settings has the following structure First the program options as defined in Programming Reference Sections 2 4 and 6 8 are listed in alphabetical order for easy reference Note that all options not appearing explicitly in the file have their default values Next the parameter values and associated flags bounds and constraint values are liste
76. une duet 33 Distributing Class Exercises A RR ey EE des 33 Maintaining and Upgrading 34 Appendix G Using Empirical Distribution Functions 35 Creating IDE S etus p Lec SE Poet ed 35 Using EDFs for AMICON edes eie tense eus 36 EDE File FOrmgt ostia Ui NR 36 Appendix Running TSM in the Condor environment 38 Installing TSM for Cond Ot eee teret d e ies eerie era 38 How Condor Works aes Ge cas te ticas t goes 39 Command Lime Interactions 40 Special Settings sicrie eicere rece pale n disc nid 40 Appendix I Tables for Nonstandard Tests 41 2 James Davidson 2015 Appendix A Installation Windows This version of the program running under Ox 7 is for Microsoft Windows 9 x Me NT4 2000 XP Vista 7 8 and 10 The following additional software components are needed to run TSM in GUI mode 1 Ox 7 10 Professional or Console or later version from http www doornik com download html Ox Console is freeware for academic use OxEdit and GiveWin are useful but non essential accessories under Windows The Professional version is neede to run 64 bit TSM 2 The Java Runtime Environment JRE freeware from http java com 3 GnuPlot fro
77. walls and virus checkers To check if this is your problem completely uninstall any suspect programs and re boot your system before trying again 3 Ifthere is a problem displaying help files and toolbar icons first make sure that the help files are located in the folder ox home packages tsmod4 help so that the program can find them Toolbar icon files with extensions gif are contained in ox home packages tsmod4 swicons Ifa problem remains or graphics files fail to display open the file t smgui41 h in a text editor and edit the line Get HomeDirectory On most systems the Ox command oxfilename 1 works OK but it might be necessary to type the path explicitly using DOS short names Don t forget to enclose the path in double quotes and use NV in place of the DOS V TSM under Linux and Mac OS X TSM for Ox 7 is available only for Windows This may change in future but meanwhile it can be run successfully on other platforms using the free Wine package see http www winehq org This section describes the implementation on an Ubuntu 13 04 system 7 James Davidson 2015 Use the Ubuntu Software Centre look under Uncategorized to install Wine and also Q4Wine a utility that implements versions of standard Windows components including the Explorer Notepad and DOS console Use Q4Wine Explorer to access the Wine emulated Windows directory structure including a C drive with Program Files and U
78. y is to create additional batch files to point to different Ox start up files Any file with the ox extension containing the first line import lt packages tsmod4 tsmgui4 gt can be used to launch TSM Batch files can be located anywhere convenient in the Start in directory for instance Use the file t smod_run bat as a template Create shortcuts to the new files as described above Installing Gnuplot TSM uses the public domain Gnuplot system for its graphics The Gnuplot version 4 2 6 executable wgnuplot exe is installed automatically with TSM The current version of the package Gnuplot 4 6 3 features a number of useful enhancements It is too large to distribute in this way but can be installed separately Download the Gnuplot installation package from http www gnuplot info If installed this version will be used automatically in place of the distributed version In case the location of the installed package is different from the default edit the Get_GnuPlotExe function in 6 James Davidson 2015 tsmgui4 h to supply the correct path Warning Upgrading your installation is essential Earlier versions of Gnuplot are not compatible To force the use of the distributed version go to Options General Special Settings and double click Plot with Gnuplot 4 2 Upgrading To upgrade an existing installation run the installation program as usual The existing file locations are retained unless edited during the setup

Download Pdf Manuals

Related Search

Related Contents

Makita HM0871C Power Hammer User Manual  USER MANUAL - Sky Elevator    ASUS T100TAF User's Manual    - Autodesk  取扱説明書  広報あつぎ 第1151号（1月15日発行）（PDF形式：2587KB）  FractionCollectorFrac-920and Frac-950  netBRICK NB 100 - Gateway-Geräte (Benutzerhandbuch)  

Copyright © All rights reserved. Failed to retrieve file