JDemetra+ in Luxembourg`s Quarterly National Accounts
Contents
1. 3 2 z 0 Normal Decomposition X11 shows the different tables produced by X11 Below you can see an A Table as an example 34 1 1 aat aas aai 1 1995 123 226 0 976 1 1 1 1 1995 160 681 1 1 1 1 1 1995 159 645 1 022 1 1 1 1 IV 1995 164 941 0 978 1 1 1 1 1 1996 108 302 1 008 1 1 1 1 1996 152 272 1 1 1 1 1 1996 146 078 0 978 1 1 1 1 1996 158 942 1 029 1 1 1 1 1 1997 113 196 0 97 1 1 1 1 1997 152 432 1 1 1 1 1 1997 150 223 1 029 1 1 1 1 A Table preliminary estimation of extreme values and calendar effects B Table preliminary estimation of time series components C Table final estimation of extreme values and calendar effects D Table final estimation of the different components E Table components modified for large extreme values 3 Quality measures 5 Summary shows the quality of the seasonal adjustment The values of the M statistics belong to 0 3 but must be 1 for a good quality Q and Q without M2 is a composite indicator calculated from M statistics They also have to be 1 for a good result Final filters Seasonal filter 3x5 Trend fiter 5 terms Henderson moving average The relative contribution of the irregular over three months span The relative contribution of the irregular component te the stationary portion of the variance The amount of period to period change in the irregular component as compared to2. Out of sample test The model 15 re estimated on linearized series for a given number lt real number of observations x of first observations and x quarter ahead forecasts and the Mean and MSE comparison between MS of forecast errors and MS of residuals are computed Regressors shows the table with all the deterministic regressors used within the RegArima part trading days leap year effect outliers Easter effect ramps intervention variables and user defined variables with their type date and value 21 Monday Tuesday Wednesday Thursday Friday Saturday 1 2008 1 1995 1995 1995 IV 1995 1 1996 11 1996 1996 IV 1996 1 1997 1997 1997 1 1997 1 1998 1998 1998 1998 1 1999 1999 1999 IV 1889 I 2000 II 2000 2000 IV 2000 i i i _ Gc i c i c i i p C i i i i i i p i i Arima Here you will find a spectral plot with peaks at frequencies Pi 2 and Pi due to the seasonality of the quarterly data The estimated coefficients of the parameters of the RegArima model are shown as well regular and seasonal AR and MA 28 RegArima model 0 1 1 0 1 1 Polynomials regular 1 00000 0 312370 B seasonal 1 00000 0 576669 5 Pre adjustment series shows a table with the series estimated by the RegArima
3. 0 978 1 029 0 97 1 029 0 973 1 018 0 979 0 998 0 999 0 998 1 018 1 008 1 1 022 0 978 H e HR Residuals shows the graph and the table of the residuals from the model 1997 07 Qi 1 of 0 E AP 1229 07 2001 07 2003 07 02 93 94 0 078 0 016 0 019 0 09 0 049 0 001 0 105 0 056 0 097 0 018 0 035 0 007 0 069 0 029 0 009 0 016 0 039 0 011 0 193 0 018 0 016 0 036 0 04 0 078 0 026 0 054 0 142 30 2011 07 gt H e e HH H HR oH oH FR oH HR HR m Statistics Analysis of the residuals Summary Here you will find 4 different tests on the residuals normality independence randomness and linearity The Portmanteau tests of Ljung Box and Box Pierce are used for the computation of tests on autocorrelation of the residuals Analysis of the residuals 1 Normality of the residuals Mean Skewness Kurtosis Normality 2 Independence of the residuals Ljung Box 165 16 Ljung Box on seasonality 2 Box Pierce on seasonality 3 Durbin Watson statistic 1 9598 3 Randomness of the residuals Runs around the mean number Runs around the mean length Up and Down runs number Up and Down runs length 4 Linearity of the residuals Ljung Box squared re
4. 5 TRD21 TRD31 RSA5C Concurrent Valid Severe E COPOS TRBi1G TRP119 RSA5C Concurrent Valid Severe E 5 TRB1G VL RSA5c Concurrent Valid Severe E COPOS TRB1G RSA5c Concurrent Valid Severe E 5 Export OR R5A5C Concurrent Valid Severe COPOS DepNat RSA5c Concurrent Valid Severe E 5 B1G_64 RSA5c Concurrent Valid Severe E 5 119 RSA5c Concurrent Valid Severe IE COPOS TRP53 RSA5c Concurrent Valid Uncertain E COPOS 52 53 RSA5c Concurrent Valid Uncertain iE COPOS TRP51 PI64 RSA5C Concurrent Valid Uncertain E 5 STR OR RSA5c Concurrent Valid Uncertain iE COPOS RSAbC Concurrent Valid Uncertain E COPOS 65 RSA5c Concurrent Valid Uncertain E COPOS ressources RSASC Concurrent Valid Good E 5 TRP72hClux RSASC Concurrent Valid Good The seasonally adjusted series can be of 6 different Quality types Good Uncertain Bad no logical error but bad quality of the results however they could be used Severe no logical error but SAProcessing should be rejected for statistical reasons Error error in the results SAProcessing should be rejected or e Missing unprocessed test meaningless test failure in execution of the test f ex for series containing only the values Empty series with no values at all will be dropped in the seasonal adjustment procedure and in the output The series with Severe Quality or Error must b
5. Cancelation limit 0 1 Final unit root limit 0 88 HR initial Initial unit root limit 1 041666666666666 Reduce 0 14236 Reduce 5E 1 0125582278481011 Unit root limit 1 05 OUTLIERS 15 enabled Detection span Use default critical value Critical value 4 Additive Level shift Transitory Seasonal TC rate 0 7 Method AddOne L5 Run X11 Mode Undefined Usually this window 15 left untouched However in the case of a detailed analysis for a series with Severe Quality the parameters in this window can be adapted in order to improve the quality of the adjustment You should also check then the drop down menu down left which gives you a detailed analysis of the adjustment performed We will call this menu Results Menu in the following Note that in this document we consider the Spec Window and the Results Menu only for an x13 adjustment They are different for a tramoseats adjustment 2 Specifications 5 8 Table 1 Pre processing El Forecasts 07 3 Regressors B o Decomposition X11 p E A Table L B Table C Table D Table E Table E Quality measures i s Summary i Details i Benchmarking E o Diagnostics Seasonality tests s Spectral analysis Residuals 1 Irregular ie 5 5a series stationary E 5 Sliding spans i M Seasonal Trading days ETET 5 series cha
6. 440 1055 439 7348 446 7614 428 9206 434 696 441 1851 446 9187 460 8053 462 99 464 5081 473 5638 477 3336 477 6647 485 2276 292 8037 305 9682 315 0028 307 1513 315 5091 326 2934 321 9755 353 5518 310 9243 330 1706 347 1577 426 7554 361 2161 386 5561 389 8154 432 0958 382 6797 398 8457 393 8307 641 802 635 6894 628 9311 639 9279 674 7924 668 8271 664 3549 688 7479 699 2924 701 8176 694 7839 729 2969 724 1206 715 4421 709 9388 747 4913 785 5036 772 7704 763 3169 87 12002 90 29827 92 52722 100 372 92 52474 96 20988 92 61128 102 2503 95 27529 94 19889 95 34453 105 116 100 9321 103 0192 99 33153 110 6017 99 80171 103 5825 102 6339 4203 991 4262 076 4175 252 4538 746 4234 135 4446 242 4284 302 4467 318 4402 92 4637 448 4558 342 4734 826 4676 07 4926 133 4869 656 5031 174 5114 366 5336 409 5263 231 load this type of Excel file right click on the Spreadsheets icon in the Providers Window then select Open Open Paste Configure Then browse for the folder using the l button on the right under Spreadsheet file TRP119 S The folder will then appear the tree under Spreadsheets Workspece window _ ODBC DSNs SDMX files 5 95 1 CC coros FF TRB1G
7. New Workspace Open Workspace Open Recent Workspace Window mai Save Workspace Save Workspace nts ESA95_1 CC Add star Open recent 111 Exit Upon saving the Workspace JDemetra creates automatically an xml file of the same name than the Workspace and containing all the informations of this one Besides a folder of the same name than the Workspace with the sub folders Calendars SAProcessing X13Doc and X13Spec or TramoSeatsDoc TramoSeatsSpec if you have used a tramoseats specification 1s created To open a Workspace previously saved go to the menu File gt Open Workspace or Open Recent Workspace and select the xml file NbDemetra 120 New Workspace Open Workspace Save Workspace Save Workspace nts ESA95 1 CC Add star n Exit Generating output To export the results go to the menu SAProcessing 1 gt Output 46 Core O File Statistical methods SAProcessing 1 View Tools V Default specification Start Refresh Accept ial E Spreadsheets C WsersWwelte Edit coros Clear selection TRB1G Specification IHE TRB Output Priority Initial Order Report The following type of files can be exported with JDemetra Csv Comma separated values Csv Matrix Excel and xt 47 Batch output
8. Sood 0 990 regarima residuals normality Bad 0 005 independence Good 0 391 spectral td peaks Good 0 582 spectral seas peaks Uncertain 0 030 visual spectral analysis spectral seas peaks Bad 0 000 spectral td peaks Good 0 000 basic checks definition Good 0 000 annual totals Uncertain 0 012 24 Beneath are displayed 2 charts the same chart as you can find under Chart and S I ratio chart oy otosa Chart shows the graphs of the original seasonally adjusted and trend series 7 Table shows the data of the original seasonally adjusted sa and trend series t and the seasonal s and irregular 1 factors S I ratio chart is very useful for analysing the development of the seasonal pattern 1 e to detect unstable or moving seasonal factors The S I ratios dots in the graph are calculated as the ratio from the original series to the trend so they present an estimate of the detrended series The blues curves in the graph represent the final seasonal factors and the red straight lines represent the average mean factor for each quarter Pre processing Summary Here you will find again the summary of the Pre Processing that you have already seen under Main results Pre processing RegArima Summary Final model Here you will find an estimation of the model by Exact Likelihood Estimation Akaike Information Criterion AIC small sample size corrected version of Akaik
9. TestStand Workspace Txt files Text files USBC files of the x13 type Xml files Extensible Markup Language Let s consider f ex an Excel input file This file needs to have the following structure The dates have to be put either in the first column column A or the first row row 1 in contrast of the titles The titles of the series have to be put either the first row row 1 or in the first column column A in contrast to the dates Values must be entered from cell B2 on Empty cells will be treated as missing values Series that contain only zeros will lead to empty series in the output Blank series will be suppressed in the output Note that the first cell cell A1 of the Excel input file can contain either nothing or anything that you like It is no more vital that the first cell is empty like in the net version of Demetra 0 C Gale ESA95 1 05 5 Insert Page Layout Formulas Data Review View Developer Add Ins Ferramenta amp Cut Calibri 1 A Wrap Text General da Copy ma E lt Format Painter Formatting as Table Clipboard Alignment Number L TRB1G TRB1G VB TRB1G VC TRB1G VF TRB1G VG TRB1G VJ VK VL TRB1G TRB1G VO VR TRB1G a janv 95 avr 95 juil 95 oct 95 janv 96 avr 96 juil
10. the amount of period period change in the trend cycle The amount of autocorrelation in the irregular as described by the average duration of run The number of periods it takes the change in the trend cycle to surpass the amount of change in the irregular The amount of year to year change in the irregular as compared to the amount of year to year change in the seasonal The amount of moving seasonality present relative to the amount of stable seasonality The size of the fluctuations in the seasonal component throughout the whole series The average linear movement in the seasonal component throughout the whole series The size of the fluctuations in the seasonal component in the recent years The average linear movement in the seasonal component in the recent years 5 Details 35 H H e e P TD amp H Mod O 0 65 2 35 18 84 0 87 224 1488 0 95 180 2257 0 99 210 9 86 Relative contributions to the variance of the percent cha in the components of the original series TOSH Total Ratio 135 10000 102 99 2 16 100 00 10509 049 10000 101 65 4 11 100 00 110 19 nents to the stationa rion of the variance in the origi 54 78 47 28 235 0 93 Total 105 57 Autocorrelation of the irregular 1 0 296 2 0 073 0110 36 Benchmarking displays the graphs and tables of the original and the benchmarked series and a graph and table of the differences between these 2 series Maximum min
11. 519 82875 7 7067 7 7884 8 2544 8 2862 11 8788 12 2982 13 9062 14 0878 14 1544 P Value 0 5248 0 7842 0 8939 0 2028 0 2792 0 2604 0 3536 0 4090 0 5056 0 2934 0 3417 0 3067 0 3670 0 4383 Ljung Box and Box Pierce tests on seasonal residuals Autocorrelation Ljung Box test P alue Box Pierce test F Value deviation 0 0607 0 1222 0 1455 0 1222 1 9304 0 1647 1 5553 0 1968 3 Randomness Runs around the mean Humber of values above the central line 33 Humber of values below the central line 34 Runs 37 Test Value Distribution Number 0 6175 0 5389 Normal 0 00 1 00 Length 10 3075 1 0000 Chiz 67 Up and down runs 42 Test P Value Distribution Number 0 605 0 4931 Hormal 0 00 1 00 Length 1 0000 Chi2 66 Ljung Box and Box Pierce tests on square residuals Autocorrelation Ljung Box test Box Pierce test P Value 0 0358 0 1 0 0957 0 1 0 0643 0 1 1 0409 0 9775 0 3228 0 0669 0 1227 1 3894 12775 0 5280 0 0223 0 1222 1 4084 1 3108 0 7266 0 1216 0 122 2 5265 2 3009 0 6606 0 09668 01222 3 1067 28055 0 7299 0 1258 0 1227 4 3474 3 8553 0 8948 0 0951 0 1222 5 0679 4 4720 0 7241 0 0743 0 1 5 5159 4 8421 0 7743 0 0839 0 1 6 0875 5 3141 0 8061 0 0297 0 1222 6 1718 53733 0 8549 0 1029 0 1227 7 0782 6 0827 0 8578 0 3233 0 1222 16 1976 13 0876 0 3627 0 0195 0 1227 16 2317 13 1132 0 4391 0 1073 0 1222 17 2746 13 8840 0 4554 ca j Ch on de
12. 96 oct 96 janv 97 avr 97 juil 97 oct 97 janv 98 avr 98 juil 98 oct 98 janv 99 avr 99 juil 99 43 56881 50 14105 44 49519 39 51443 39 81469 44 92717 44 67542 42 54846 32 49365 36 79812 37 86218 36 82341 37 46451 44 28427 43 47094 40 5965 46 00193 51 35958 48 42593 530 5407 552 5262 506 2069 511 7684 514 3278 541 6562 506 7003 534 3376 534 3193 571 6134 536 4287 567 4098 589 6934 596 1351 567 9353 569 1145 597 0763 628 014 610 3862 450 8463 473 0992 432 3406 433 2456 431 1662 459 1421 430 2271 452 75 443 7766 487 0918 458 3869 483 5545 499 6697 505 9252 482 7236 480 7319 502 5912 537 9709 527 2669 254 1733 321 8997 264 3995 309 104 213 2065 324 4493 273 5801 299 9188 253 895 342 2137 289 4452 319 4039 299 8895 364 5304 313 1493 316 5808 334 0607 381 6867 315 8417 738 3087 784 8044 763 3856 809 1277 733 4923 792 2381 778 9013 808 7565 755 9407 829 3841 819 5113 890 879 817 0665 884 1754 887 8697 941 439 867 837 939 2593 925 6758 179 8683 183 4238 176 7612 205 4618 174 3049 185 9337 186 7726 220 9724 180 6665 187 7389 194 5104 218 1883 213 6167 236 6437 244 7058 269 5093 273 1351 314 9643 322 0376 1025 696 895 7586 950 341 1172 069 1068 712 1042 214 998 1165 986 9223 1132 379 1124 859 1125 79 1025 001 1076 232 1134 159 1159 364 1131 683 1266 726 1259 88 1298 191 423 4865 439 9349 451 4473 470 7371 437 4289
13. CX matrix Excel Txt Usually Excel data is exported Select Excel and then double click on the Excel icon on the left Now you can define the folder where your output file will be saved to and the name of the exported file under the node gt Location gt folder and gt File Name If you want to get the same output structure as in the input file you should choose under the node gt Layout gt layout ByComponent and tick gt verticalOrientation For Quarterly National Accounts the data sent to Eurostat include the raw original data y the seasonally adjusted data sa and the working day corrected data ycal So you should select these under the node gt Content gt series Note that many other series can be exported like forecasts trend irregular etc 48 Batch output I Location folder C Users velter D m File Name ESA95_1_COPOS Layout layout ByComponent verticalOrientation Content ENSEM The output option will create by default a folder of the same name than the Workspace f ex Workspace 1 at the place defined under folder 1n the example above C Users velter D In this folder another folder of the name of the processing f ex SAProcessing 1 which contains the series in the above example ESA95 1 COPOS can be found Beware that in contrast to Demetra 2 03 JDemetra drops empty columns from the input in the output and so if you have
14. Co Distribution shows the autocorrelogram and the partial autocorrelogram and a histogram of the residuals estimated from RegArima model The grey dotted lines in the ACG and PACG represent the confidence intervals As long as the partial autocorrelation coefficient is in the confidence interval everything is all right However you should pay special attention to the series where the partial autocorrelation coefficient 15 greater in absolute value than the limits of the confidence interval In fact the ACG and PACG help to identify the model o Ifthe PACF cuts off sharply after lag p and the ACF declines in geometric progression from its highest value at lag p then the series can be modelled by an AR p model 33 Ifthe cuts off sharply after lag and the declines geometric progression from its highest value at lag q then the series can be modelled by an MA q model If the series is not stationary it needs to be differenced first parameter d in an ARIMA p d q model indicates how often the series had to be differenced until it was stationary The histogram shows if the residuals are normally distributed Autocorrelations 0 25 on 0 45 0 18 20 10 0 05 0 4 0 00 0 05 0 35 0 10 0 20 0 3 D 25 0 30 uu 0 6 12 18 24 30 36 Partial autocorrelations 0 20 0 2 0 15 0 10 0 15 0 00 0 05 0 10 0 15 0 05 A 0 25
15. Intervention variables Ramps User defined variables E ARIMA Automatic Accept Default CheckMu LjungBox limit 0 95 Mixed ArmaLimit 1 Cancelation limit 0 1 Final unit root limit 0 88 HR initial Initial unit root limit 67 Reduce CV Reduce SE 1 01255827 784381011 Unit root limit 1 05 OUTLIERS 15 enabled Detection span Use default critical value Critical value 4 Additive Level shift Transitory Seasonal TC rate 0 7 Method AddOne LS Run 0 X11 Mode Undefined First of all we can change the series span This could be useful f ex in the case of a big financial crisis where you would like to split the series in 2 parts one before and one after the crisis However the model is supposed to be robust enough to deal with this without having to split the series and another problem could then be that you wouldn t have enough data for a good model for one part of the series 42 SERIES Series span type 1 ESTIMATE Model span Tolerance E TRANSFOR function AIC difference Adjust E REGRESSION Excluding Another option 15 to restrict the model span which is the data span on which your model will be based This could be interesting f ex 1n the case of Quarterly National Accounts In October each year annual data becomes available and this data 1s supposed to be less volatile than quarterly data so you could fix your model during the whole year to this The only problem is then that you ca
16. JDemetra Luxembourg s Quarterly National Accounts Overview This document describes in detail how seasonally and calendar adjusted Quarterly National Accounts are compiled with the help of JDemetra at STATEC National Statistical Institute of Luxembourg As soon as the production version of JDemetra 15 available probably in November 2013 Quarterly National Accounts Employment and Turnover will be seasonally calendar adjusted forecasted using JDemetra As an example let s take a look at the procedure for compiling adjusted Quarterly National Accounts First a new JDemetra session needs to be opened KeePass 2 k Adobe Reader X 9 MbDemetra 1 2 0 a p Paint Xa Mozilla Firefox k Demetra UI o REP2000 V9 0 Enterprise Guide 4 d 181 Visual Builder IBM SPSS Statistics 20 Demetra b k All Programs Upon opening JDemetra presents 2 default windows the Providers Window and the Workspace Window In the Providers Window the input data will be charged It can be a MySql database ODBC DSN an Oracle database a SDMX file a Spreadsheet a TSW a Txt file an USBC or an Xml file Most data used at STATEC are in Excel 2010 format In the Workspace window we can select a seasonal adjustment model of the family tramoseats or x13 NbDemetra 1 2 0 File Statistical methods View Tools Window Help M ik Chart amp arid m 3
17. TRB1G VF TRB1G VG I TRB1G VL Hf 16 VM EE mevo Hf TRB1G VR_U 16 119 EE TRD21 TRD31 E DOTRB18G EE 071 amp In the drop down structure you will see the path of the folder the name of the sheet 05 and the name of the series in the sheet TRBIG VA TRBIG VB E TRBIG VC TRBIG VF etc If you want to close a file right click on the file path in the Providers Window then select Close 10 eee TT Once you have uploaded the series 1 the Providers Window you can start with the seasonal adjustment procedure You can chose between the following specifications in 5 Add star TRB1G V TRBi1G VB E TRB1G VC TRBiG VF Close TRBi1G VG I Reload TRB1G VJ TRB1G VK TRB1G VL Copy TRB1G VM TRBi1G VO TRB1G VR TRB1G TRP 1189 TRB 1G TRP119 TRD21 TRD31 TRE 104 021 Edit Clone Rename as xml Mail as xml Seasonal adjustment with JDemetra the Workspace Window 11 Providers Window Workspace wi 5 Workspace Cours 0 NACE Rev E Modelling Seasonal adjustment tramoseats Fl RSA RS5A1 RSA2 RSA4 RSAS E X11 S RSA1 RSA2c RSA4c UN R amp AB5c E a documents amp I multi documents E Calendars Description of the different specifi
18. Workspace 1 Modelling 5 Seasonal adjustment Gu Utilities Plugins First you should make sure that all the installed plugins in JDemetra are activated To do so go to gt Tools gt Plugins Tools Window Help Container d Spectral analysis F Differencing Plugins Options Check that all of the 11 installed plugins are active sign after the name of the plugin Category NbDemetra Core eu europa ec joinup NbDemetra UI eu europa ec joinup Version 1 2 0 NbDemetra 5A Advanced eu europa ec joinup Source NbDemetra 1 2 0 NbDemetra Branding eu europa ec joinup NbDemetra 5A LI el Plugin Description NbDemetra Spreadsheet i Wrapper module for Demetra core libraries NbDemetra Common NbDemetra JDBC NbDemetra ODBC NbDemetra 50 RCP Platform RCP Platform NbDemetra Core a If you want to add self made plugins go to the flag Downloaded and click on the Add Plugins button Browse for the plugins you want to insert These are now added to the installed plugins and need to be activated Install Note that the plugins are essential to JDemetra They are in fact the core element JDemetra and main difference to Demetra net version Charging the national calendar As JDemetra doesn t have any predefined calendars like in old Demetra 2 3 the national calendar needs to
19. be charged manually To do so right click on the Calendars icon in the Workspace Window and select gt Add Calendar and then gt National NoDemetra File Statistical methods View Tools Window Help Workspace_1 Modelling s Seasonal adjustment Ej u Utilities 5 P Calenda Defa Add Calendar Mational ERE Variables Chained Import Composite In the following window give the calendar a name f ex Luxembourg Upon clicking on the sign you have the choice between Fixed Easter related Fixed Week and Special Day Luxembourg has e 2 Fixed holidays National Day 23 June and St Stephen s day 26 December These have to be inserted manually in the calendar Nevertheless you can add St Stephen s day also as Special Day Christmas with Offset 1 which means the day after Christmas e no Easter Related Days as this effect is not so important to Luxembourg s economy In fact Christmas is a Holiday where much more revenue from food presents and decoration is made Easter Related Days determine the days before and or after Easter where an Easter effect 15 present As tramoseats was developed in Spain where the Easter effect 1s very important special attention has been given to this holiday in the program e Fixed Weeks as this effect 15 not so important to Luxembourg s economy except for the constru
20. cations for tramoseats RSAO airline model 0 1 1 0 1 1 RSAI log test outlier detection airline model RSA2 log test test on the presence of working days 1 parameter working day or not Easter effect outlier detection airline model RSA3 log test outlier detection automatic identification of the model RSA4 log test test on the presence of working days 1 parameter Easter effect outlier detection automatic model identification RSAS log test test on the presence of trading days 6 parameters Monday Tuesday etc Easter effect outlier detection automatic model identification and for x13 without any pre adjustment RSAI log test outlier detection airline model 12 RSA2C log test test on the presence of working days 1 parameter working day Easter effect outlier detection airline model pre adjustement for leap year if logarithmic transformation has been used RSA3 log test outlier detection automatic model identification RSA4c log test test on the presence of working days 1 parameter Easter effect outlier detection automatic model identification pre adjustement for leap year if logarithmic transformation has been used RSASc log test test on the presence of trading days 6 parameters Monday Tuesday etc Easter effect outlier detection automatic model identification pre adjustement for leap year if logarithmic transformation has been used For the c
21. ction branch Fixed Weeks can f ex be used when the activity of a whole branch 15 low for one or more weeks like in Luxembourg during the collective leave period in 2012 27 07 19 08 and 22 12 09 01 e 10 Special Days New Year Easter Easter Monday Ascension Pentecost Whit Monday Day Assumption All Saints Day and Christmas These have to be selected on the right side of the window under gt Special Day gt Day event Note that the Special Days have a Weight function under gt Event gt Weight where you can assign the impact of a holiday on the national economy f ex 0 5 for half day off and a given percentage if only part of the economy doesn t work They also have an Offset function under gt Special Day gt Offset where you can determine the days before and or after the Special Day where an economic effect is present f ex if all or some branches of the economy don t work the day before or after a Special Day In Luxembourg the bank sector has a day off on Good Friday so you could treat this as a Special Day Good Friday with Weight 0 17 as the bank sector contributes to 1796 to the GDP of Luxembourg Another example is Christmas Eve a day on which most people work only half days so we could consider Christmas Eve as Special Day Christmas with Offset 1 and Weight 0 5 In practice half days and days where only part of the economy have a day off have no effect on the Calendar adjustment So this is o
22. e Multiplicative LogAdditive whether you want JDemetra to compute forecasts the lower and upper sigma boundaries for the detection of extreme values the kind of seasonal filter seasonal moving average used to estimate the seasonal factors whether the length of the Henderson filter used in the estimation of the trend is detected automatically or if not the length of it X11 Made Undefined Seasonal component v Use forecasts v LSigma 1 5 USigma 2 5 Seasonal filter Msr Details on seasonal fil Automatic henderson filter 57 Henderson filter 13 In the Benchmarking node you can enable or disable benchmarking and fix the target to the original or calendar adjusted data which means that you can determine whether the sum of quarters of the seasonally and calendar adjusted data equals the sum of quarters of the original data or the sum of quarters of the calendar adjusted data Benchmarking makes sense if the annual totals the Diagnostics basic checks bad BENCHMARKING Is enabled Lambda Note that in the lower part of the Spec Window you can find a small window with the description of each element of the Spec Window Before exporting the results you need to save the Workspace To do so go to the menu File gt Save Workspace As and save it under the name and place you want 45 NbDemetra 1 2 0 Statistical methods View Tools Window Help
23. e Information Criterion AICC and Bayesian Information Criterion BIC are used to select the best model The results of the outlier detection are also shown JDemetra detects 3 types of outliers TC Transitory Change LS Level Shift and AO Additive Outlier 25 Summary Estimation span 1995 v 2012 T2 observations Series has been log transformed Series has been corrected for leap year Trading days effects 6 variables easter effect 1 detected outlier Final model Likelihood statistics Humber of effective observations 67 Humber of estimated parameters 10 Loglikelinood 92 49930337363443 Transformation adjustment 355 70672599456695 Adjusted loglikelihoad 263 2074236210325 Standard error of the regression ML estimate 0 058200582732019376 AIC 545 414847242065 AICC 550 3434136706354 BIC corrected for length 5 122910357095336 Scores at the solution 0 003019 0 004596 Arima model 0 1 1 0 1 1 Coefficents T Stat Theta 1 0 3124 2 50 BTheta 1 0 8769 13 27 ression model Trading days Coefficents Monday 0 0222 Tuesday 0 0281 Wednesday 0 0209 Thursday 0 0016 Friday 0 0175 Saturday 0 0217 Joint F Test 3 89 0 0025 Qutliers Coefficents T Stat TC L 2008 0 2228 482 Forecasts shows forecasts of the original series y_f and their confidence interval 26 2010 01 2010 07 2011 01 2011 07 2012 01 2012 07 2013 01 2013 07
24. e treated case by case If you want to accept a series with Severe Quality or Error select the series right click and select Accept COPOS 119 ee Valid Severe Refresh F Valid E 5 TRB1G VL rent Severe E COPOS rent Valid Severe E COPOS Export OR Edit p rent Valid Severe E COPOS DepNat T rent Valid Severe Specification E 5 64 l irent Valid Severe 19 This option only be used for unimportant series that do not affect GDP are not taken into account in the results analysis By clicking on the Specifications button on the top right in the SAProcessing 1 window a window appears down right where you can change many parameters for each series f ex the series span the model span the calendar the ARIMA model chosen the level of significance of the tests the boundaries for the detection of extreme values whether or not to benchmark etc We will call this window Spec Window in the following 20 SERIES gt Series span ESTIMATE Model span Tolerance 0 0000001 TRANSFORMATION function Auto AIC difference 2 None LEGRESSION Calendar Pre specified outliers Intervention variables Ramps User defined variables E ARIMA Automatic Accept Default CheckMu LjungBox limit 0 95 Mixed ArmaLimit 1
25. find a sliding spans statistic a cumulative frequency distribution of the sliding spans statistic and a table with the breakdowns of unstable factors and average maximum percentage differences across spans The sliding spans statistic calculates the maximum percentage difference for each quarter The cumulative frequency distribution of the sliding spans statistic uses a frequency polygon where the x axis shows the values of the sliding spans and the y axis the frequency in percentage of each class interval In the third panel the abnormal values indicate the percentage of values that do not meet the sliding spans condition If the sliding spans Statistic reveals a large number of unstable estimates you should think about changing the model specification 40 3 ggg 2000 0 20110 01 2011 01 Distribution 2012 01 20 00 00 For a more detailed description of the tests you should have a look at the Demetra User Manuel or to the JDemetra User Manuel as soon as it will be available Any Time Series Manuel will also give you the information needed on these tests Let s now consider how we can change a Series with a Severe quality output To do so we have to perform changes in the Spec Window 4 SERIES Series span ESTIMATE Model span Tolerance 0 0000001 TRANSFORMATION function Auto AIC difference 2 Adjust None REGRESSION Calendar Pre specified outliers
26. ied o Intervention v Ramps User defined The most important thing in the ARIMA node 15 the fact that you can enable or disable Automatic modelling Sometimes JDemetra tries to force a seasonal model on a series that is definitely not seasonal already from a graphical point of view and you will get a severe quality output In this case you can chose to model the series yourself and set the BP BD and BQ parameters to 0 so that your ARIMA model will not be seasonal any more Usually you get a model of much better quality then 1 1 ARIMA Accept Default CheckMu LjungBox limit Mixed ArmaLimit Cancelation limit Final unit root li HR initial Initial unit root Reduce CV Reduce SE Unit root limit ia 0 95 ARIMA Automatic 1 Mean 0 1 D 0 88 o theta 1 0416666666 BP 0 142536 1 0125582278 1 05 btheta n C EI For the outliers you specify amongst others the detection span for the outliers whether a default critical value is used or if not the value of this one and the type of outliers detected additive level shift transitory and seasonal 44 jj OUTLIERS Is enabled Detection span Use default critical value Critical value 4 Additive Level shift Transitory Seasonal E TC rate 0 7 Method AddOne 15 0 the X11 node you can specify the decomposition mode Undefined Additiv
27. igma 1 5 x11 usigma 25 x11 trendma 0 x11 fcasts 1 benchmarking rho 1 0 benchmarking enabled false benchmarking bias Hone benchmarking target Original benchmarking lambda 1 0 Main results Pre processing RegArima Summary Here you will find amongst others the estimation span the number of observations if the 23 series have been log transformed if trading days have been detected if an Easter effect has been detected and the number of outliers detected COP05 TRP51 Summary Estimation span H1995 v 2012 T2 observations series has been log transformed series has been corrected for leap year Trading days effects variables easter effect 1 detected outlier Diagnostics Here you will find amongst others the quality of the adjustment Good Uncertain Bad Severe Error or Missing if residual seasonality is still present in the series the m Statistics to evaluate the quality of the seasonal adjustment the regarima residuals the visual spectral analysis to identify spectral peaks in seasonal and trading day components and the annual totals to compare the annual totals of the original series with those of the seasonally adjusted series Diagnostics summary Uncertain residual seasonality on Sa Good 0 351 on sa last 3 years Good 0 959 irregular Good 0 420 m statistics Good 0 154 q without m2 Good 0 170 out of sample mean Good 0 933
28. imum average and standard deviation of the differences are calculated as well original result Differences 1 1995 155 844 153 918 0 012 1995 155 635 153 975 0 011 III 1995 155 624 154 492 0 007 IV 1995 146 428 146 108 0 002 1 1996 132 077 132 671 0 004 1996 147 738 148 991 0 008 III 1996 149 217 150 627 0 009 IV 1996 142 257 143 306 0 007 1 1997 142 793 143 134 0 002 1997 148 323 148 133 0 001 III 1997 146 929 146 402 0 004 IV 1997 154 851 154 148 0 005 1 1998 161 589 160 931 0 004 19 7 098 07 2000 07 2002 07 2004 07 2 008 07 2008 07 20 10 07 2012 07 1998 157 941 157 41 0 003 1998 157 23 156 847 0 002 original result 1 1998 157 513 157 311 0 001 1 1999 154 856 154 872 0 001 5 1999 150 94 151 215 0 002 III 1999 154 16 154 758 0 004 IV 1999 149 927 150 865 0 006 Relative differences 0 01 Max 0 013731707236292579 Min 0 013880227326819372 004 07 2006 07 2008 07 2010 07 2012 07 Average 8 248643693748664 5 s uns mE m m Stdev 0 007018422150688345 Differences Diagnostics Here you will find again the diagnostics that you have already seen under Main results Diagnostics Seasonality tests Here you will find a few tests to check the presence of seasonality in your series Friedman test test for stable seasonality Kruskall Wallis test test for stable seasonality Test for the presence of seasonality assuming stability Evolutive seasona
29. in the denominator Palve 0 3557 Ho evidence of moving seasonality at the 20 per cent level Combined seasonality test Identifiable seasonality present Residual seasonality test No evidence of residual seasonality in the entire series at the 10 0 per cent level F 0 1026 Ho evidence of residual seasonality in the last 3 years at the 10 0 per cent level 2 0 0383 Spectral analysis 38 Here you will find spectral plots to check the presence of remaining seasonal and trading day effects 2 spectral plots are displayed the periodogram and the auto regressive spectrum The graphics are available for the residuals Residuals the irregular component Irregular and the seasonally adjusted series Sa series stationary The periodicity at frequency f is 27 f so we have 2 seasonal frequencies for quarterly series 1 2 and In the plots the blue grey lines correspond to the seasonal frequencies and the red lines to the trading day frequencies If we have a peak at a seasonal or a trading day frequency a better fitting model is needed A peak at a seasonal frequency means that the filter used in the decomposition is not well adapted to the series A peak at a trading day frequency means that the regression variables in the model are not well adapted or that the calendar effects are changing too much The green horizontal line denotes the 005 significance level so the peaks above this line are significant at 005 level No
30. lity test test for moving seasonality Combined seasonality test combined seasonality test using Kuskall Wallis test Test for the presence of seasonality assuming stability and Evolutive seasonality test o Residual seasonality test for the seasonally adjusted series on the entire time span and on the last 3 years O O O 0 0 37 Non parametric tests for stable seasonality Friedman test Friedman statistic 1 Distribution F 3 51 P V alue 0 0000 Stable seasonality present at the 1 per cent level Kruskall Vfallis test Kruskal Wallis statistic 65 35616438356163 Distribution Chi2 3 P Value 0 0000 Stable seasonality present at the 1 per cent level Test for the presence of seasonality assuming stability Degrees Sum of squares of Mean square freedom Between months Residual 0 05587566935969453195 60 0 2 5 275863142566647 4 Total 1 0765481302079403 71 0 0 015162649721235595 1 0406 72260030467 3 0 0 34690753612829 Value 657 575216187357 Distribution F with 3 degrees of freedom the nominator and 68 degrees of freedom in the denominator 0 0000 seasonality present at the 1 per cent level Evolutive seasonality test Degrees of freedom 0 00789292913067568378 17 0 gt 4 12054366507655 4 Sum of squares Mean square Error 0 021016302691590913 51 0 Value 1 1275848137829637 Distribution F with 17 degrees of freedom in the nominator and 51 degrees of freedom
31. ment s specifications A new window SAProcessing 1 opens now Drop the data from the Providers Window in the empty space in the SAProcessing_1 window by drag and dropping the Excel sheet icon with all the time series By default the specification will be RSA4 If you want a different specification select all the series in this window then right click and select gt Specification and chose the one you want 17 Matrix 178items 7 TS RSA4 5 TRB1G S 5 16 VB COPOS 16 V S 5 TRB1G 5 TRB1G 5 TRB1G VJ COPOS TRB1G cert coos meiewm n 444 4 S 5 Q Sj 5 TRB1G VR U 05 TRB1G 5 TRP119 i Refresh Edit Specification Lh X 135pec 1 Note that the series are not processed yet To seasonally adjust the series you need to click 9 button Seasonal adjustment is quite fast in new JDemetra and the results will be displayed immediately 18 SAProcessing 1 3 Processing Summary Matrix 178 F TS RSA4 Series Method Estimation Status Quality COPOS ControleImport RSA5c Concurrent Invalid E COPOS ControleExport RSA5c Concurrent Invalid E COPOS Import OR RSA5c Concurrent Valid Error E 5 TRP51PI53 RSA5C Concurrent Valid Severe E
32. mitted in Luxembourg s calendar Note also that for all of these holidays a starting date and an ending date can be defined under gt Event gt Start gt End This is f ex useful if a national holiday is proclaimed abolished f ex Victory in Europe Day 8 of May has been a national holiday in France from 1953 to 1959 was then abolished and became again a national holiday from 1982 on However this 1s not relevant for Luxembourg Edit National Calendar Fixed Easter Related Fixed Week Special Day Special Day Ascension Special Day Pentecost 0 Special Day WhitMonday 0 Properties gt Special Day MayDay 0 Special Day Assumption 0 Special Day AllSaintsDay 0 Special Day Christmas 0 Clicking the button will add your new calendar under the Calendars icon in Workspace window Note that Calendars can be edited cloned or removed by right clicking on the calendar s name Providers Window Workspace Wi Lj Workspace Cours 20 NACE Rev2 m Modelling s Seasonal adjustment 21 10 Utilities 7 Calendars i P Default Le Luxembourg E v Variables Edit Clone Remove Loading the data The following type of input files can be loaded in JDemetra ODBC DSNs Open Database Connectivity Data Source Name SDMX files Statistical Data and Metadata eXchange Spreadsheets Excel files TSW
33. n get higher revisions once you change the model in October of the following year El ESTIMATE 2 Model span C Tolerance f TRANSFOR 7 function AIC difference Adjust Between 2 REGRESSION Last Calendar First Pre specified Exduding Considering the transformation node you better leave this one untouched Here you can force the transformation to a Log transformation but it is better to let the program detect the transformation needed You can also change the Akaike difference and the Adjustment to Leap Year and to Length of Period but so far this has not proved to be so useful to us TRANSFOR function AIC difference Adjust REGRESSION The regression node however is more interesting Here you can define whether you want to adjust for trading days or not apply your calendar adjust for leap year adjust for Easter define the duration of Easter etc You can also define pre specified outliers intervention variables ramps and user defined variables All of these are very interesting f ex 1 you want to test the effect of a trading days adjustment different calendars adjustment for leap year or Easter and different sets of outliers intervention variables 43 ramps and user defined variables on your series REGRESSION 3 Calendar E trading in use option Default td TradingDays LeapYear aut test Remove easter in use Is e Pre Add Eas 8 Pre specif
34. nges First we will have a look at the Results Menu Specifications Here is a summary of the chosen specification with all the parameters and options 22 algorithm Seasonal adjustment x13 0 1 0 0 regarima algorthm Modelling regarima 0 1 0 0 regarima basic span regarima basic preprocess true regarima transform adjust Hone regarima transform const 0 0 regarima transform function Auto regarima transform aicdiff 2 0 regarima arima theta 1 coeff regarima arima btheta 1 coeff regarima arima mean false regarima arima d 1 regarima arima bd 1 regarima automdl enabled true regarima automdl acceptdefault false regarima automdl cancel 0 1 regarima automtadl ljungboxlimit 0 95 regarima automdl reducecwv 0 14286 regarima automdl ub2 0 88 regarima automdl ub1 1 0416666666656667 regarima outlier span regarima outlier lsrun 0 regarima outler tcrate 07 regarima outlier tc 0 0 regarima outlier method Add ne regarima outlier lz 0 0 regarima outlier ao 0 0 regarima outlier maxiter 30 regarima outlier defcwv 0 0 regarima esimate span regarima esimate tol 1 0E 7 regarima regression tradingdays option TradingDays regarima regression tradingdays test Remove regarima regression tradingdays leapyear Leap ear regarima regression tradingdays autoadjust true regarima regression tradingdays stocktd 0 regarima regression mh1 param regarima regression mh1 test Add regarima regression mh1 type Easter x11 mode Undefined x11 Is
35. ompilation of Quarterly National Accounts STATEC uses x13 with a maximum of parameters so the specification RSASc In addition the calendar for Luxembourg previously defined needs to be integrated do so clone RSA5c specification by right clicking on it and selecting Clone Workspace Wi 3 Workspace Cours 0 NACE Rev2 E 4 Modelling s Seasonal adjustment specifications tramoseats RSAO RSAL RSA2 RSA4 RSA5 LETETI xii RSA1 RSA4C RSAC ig igo ign ig ug ig ig ig ba a documents Open E 1 multi document Utilities Delete Calendars Rename 0 P Default in F Luxembour Create Document H v Variables Clone A new X13Spec 1 specification is then created 13 X11 RSA1 RSA4c RSA5c X135pec 1 Double click on the newly created specification X13Spec 1 specification and add the National Calendar for Luxembourg To do so go to the node gt Regression gt Calendar gt tradingDays gt option where you select Holidays instead of Default and gt holidays where you select your previously defined National Calendar here Luxembourg instead of Default X135pec 1 SERIES Series span ESTIMATE rj Model span Tolerance TRANSFORMATION function AIC difference Adjust REGRESSION Calendar E tradingDays option easter Pre specified outliers In
36. on the X13Spec 1 specification and choose Create Document 15 m d ri 2 0 Delete c U Utilities Delete EF n Calendars Can PO e Zim Default C Ig vi B7 Luxembourg v Variables Clone e you go via the menu Statistical methods gt Seasonal Adjustment gt Single Analysis gt X13 File View Tools Window Help Modelling Seasonal Adjustment Multi Processing e DE S Jom _Cours_30 _NACE Tank X13 B Seasonal adjustment EF Ls specifications tramoseats Structural Model Mixed Airline Generalized Airline 09 bus ig ig ig ig ig Now X13Doc 1 window opens the right where you can drop the data of only one series under Drop data here This series 15 directly seasonally adjusted and results are displayed in the window on the right 16 Specifications Main results 1 Pre processing However we usually need to seasonally adjust multitude of time series do so we only have one option Go to the menu Statistical methods gt Seasonal Adjustment gt Multi Processing gt New NbDemetra 1 2 0 Modelling a Seasonal Adjustment Multi Processing New Providers Window Workspac Single Analysis Workspace Cours 0 NACE Tools H Modelling B 5 Seasonal adjust
37. part In our tests the following series appeared Interpolated series yc Linearized series y_lin Series corrected for calendar effect ycal Deterministic component det Calendar effect cal Trading day effect tde Easter effect ee Outliers effect on the irregular component out_1 Outliers effect on the trend component out t Total outliers effect out O 0 0 Q O 0 OQ O9 20 1995 1995 11 1995 IV 1995 1 1996 1996 1996 IV 1996 I 1997 199 1997 IV 1997 160 681 159 645 164 941 108 302 152 272 146 078 168 942 113 196 152 432 150 223 175 966 135 333 161 851 151 679 127 922 154 32 150 106 179 362 121 756 147 385 142 409 191 573 126 285 160 681 156 211 168 724 107 41 152 272 149 351 164 259 115 746 152 432 146 059 179 679 132 904 161 851 154 879 183 937 128 067 154 32 150 352 176 251 120 753 147 385 139 345 195 967 126 285 160 681 156 211 168 724 107 41 152 272 149 351 115 746 152 432 146 059 179 679 132 904 161 851 154 879 183 937 128 067 154 32 150 352 176 251 120 753 147 385 139 345 195 967 0 976 1 022 0 978 1 008 0 978 1 029 0 97 1 029 0 979 1 018 0 979 0 998 0 999 i 0 998 1 018 1 008 1 1 022 0 978 0 976 1 022 0 978 1 008 0 978 1 029 0 97 1 029 0 975 1 018 0 979 0 998 0 999 1 0 998 1 018 1 008 1 1 022 0 978 out i 0 976 1 022 0 978 1 008
38. rs chosen by JDemetra for seasonal and calendar adjustment are only changed as new annual data becomes available which means that the exercise above is only repeated at annual frequencies in October For the other compilations of QNA data in January April and July the model chosen remains the same but new data is taken into account same model span new time span To do so select the previously seasonally adjusted file in the Workspace Window under the node gt Seasonal Adjustment gt multi documents by double clicking on it The corresponding SAProcessing will open then in the window on the right Select all the series in this processing right click and select gt Refresh gt Current adjustment partial The series are then seasonally adjusted based on the former parameters but integrating the new data 50 SAProcessing l amp Summary Matrix Series li 5 5 li coros TRE1G li coros TRB1G VF lj 5 TRE1G VG Bi coPo5 TRB1G li 5 TRB1G VK I coros mei vo gQ Concurrent Specification li 5 COPOS 118 Ej TSIRSA4 Refresh Current adjustment partial Accept Partial concurrent adjustment Note that you only need to update your Excel file with the new data and not to import it again in JDemetra as this is done automatically in the Providers window Appendi
39. series with no values at all in your input they will automatically be cancelled out in the output So you should use lookup formulas in any Excel files that use data from your output In the output file seasonally adjusted data can be found on the sa sheet Working day adjusted data will be on the ycal sheet whereas original data will be on the y sheet Note also that in the output file the series have been shifted one column down compared to the output of Demetra 2 03 Now a new line with the title of the adjustment sa ycal or y has been inserted at the beginning of each sheet This 1s important in the case that any macros point on the data If you haven t saved your workspace yet you should do this now Your processing in this case SAProcessing 1 will then be available for future refreshment in the Workspace Window under the node seasonal adjustment gt specifications gt multi documents gt 5 1 49 Providers Window Workspace Wi Workspace Cours JD4 NACE Rev2 H m Modelling L 1 L 1 Seasonal adjustment s specifications documents Tamoseats x13 Structural model Generalized airline model E m Mixed airline model k multi documents AProcessing 1 i u Utilities g Calendars v Variables Reopen a JDemetra processing In Luxembourg s Quarterly National Accounts paramete
40. siduals 15 Box Pierce on squared residuals 16 Details P value 0 3347 0 2720 0 0045 0 0050 P value 0 2605 0 4383 0 1547 0 0722 P value 0 5369 1 0000 1 0000 1 0000 P value 0 2418 0 4584 you will find the details on the tests from the summary above Besides the p value you can also see the autocorrelation standard deviation and distribution 31 Details 0 Statistics Sum of squares 0 2269 MSE 0 0035 Standard error 0 0581 1 Distribution Standard deviation 0 0578 Hormality tests Test Value P Value Skewness 0 3287 0 2720 Kurtosis 4 5969 0 0046 Joint test 10 5897 0 0050 2 Independence tests Ljung Box and Box Pierce tests on residuals Standard deviation 0 0015 0 1222 0 0584 0 1222 0 0062 0 1222 0 0607 0 1222 0 0432 0 1222 0 2823 0 1222 0 0708 0 1222 0 1455 0 1222 0 0298 0 1222 0 0854 0 1222 0 0218 0 1 0 2315 0 1222 0 0793 0 1222 0 1549 0 1222 0 0535 0 1222 0 0291 0 1222 Autocorrelation Oo C fo R3 i i P Value 0 3347 Ljung Box test 0 2539 0 5245 0 8635 5 7047 7 0907 8 7504 8 8214 9 4122 9 4513 13 9556 14 4944 16 5878 18 8421 16 9187 32 Distribution Hormal 0 00 0 30 Normal 3 00 0 60 Chi2 2 P M alue 0 5144 0 7693 0 8818 0 1523 0 2140 0 1881 0 2657 0 3087 0 3967 0 1750 0 2068 0 1658 0 2066 0 2605 Box Pierce test 0 2391 0 4862 0 8111 5 9
41. te that you can Copy and Print all of these graphs Penodogram Pia Auto regressive spectrum Sliding spans analyses the stability of the seasonally adjusted series These are supposed to be stable in the sense that adding or removing data points at the beginning or ending of the series doesn t affect the seasonally adjusted series very much Sliding spans allow you to detect significant changes in your series like seasonal breaks large number of outliers and fast moving seasonality A span is a range of data between 2 dates Depending on the length of the original series the sliding spans include 2 3 or 4 overlapping spans The maximum of spans 15 set to 4 and the intervals on which the spans begin to 1 year The Sliding spans summary shows the time spans used tests for seasonality for each of these time spans and means of seasonal factors for each quarter in each span 39 Time spans Span 1 from 2002 to 2009 Span 2 from 2003 to 201 Span 3 from 2004 to 2011 Span 4 from 2005 to 2012 Tests for seasonality Stable seas Kruskal Wallis Moving seas identifiable seas Means of seasonal factors 0 8504 8598 0 8553 1 0435 1 0407 0 9450 0 9524 1 1510 1 1519 more detailed analysis is given for the seasonal component Seasonal the trading days effect 1 Trading days and the seasonally adjusted series SA series changes For each of these you will
42. tervention variables 0 0000001 Auto 2 None in use Holidays Luxembourg TradingDays LeapYear Remove in use Note that if you want to benchmark all of your series it 1s easiest to do this in this window Go to the node gt Benchmarking where you check the Is enabled box Then under gt Target select either Original or CalendarAdjusted whether you want the adjusted quarterly series to be fitted to the sum of quarters of the original data or to the sum of quarters of the calendar adjusted data 14 8 X13Spec 1 5 4 Level shift Transitory Seasonal TC rate Method LS Run X11 Mode Undefined Seasonal component Use forecasts LSigma 1 5 USigma 2 5 Seasonal filter Msr Details on seasonal filters Automatic henderson filter Henderson filter 13 BENCHMARKING Is enabled Rho Original Lambda CalendarAdjusted Target Target However no benchmarking is done for Luxembourg s Quarterly National Accounts as this would distort the series patterns and lead to stronger revisions in quarterly data Now the new specification X13Spec 1 contains all the parameters of the RSA5c specification as well as the National calendar for Luxembourg If you want you can rename the newly created specification by right clicking on it and then selecting gt Rename Now everything is ready for seasonal adjustment of the series For single analysis you have 2 options e either you right click
43. x JDemetra be downloaded here http www cros portal eu content 1demetra 1ava version 1includineg source codes A User Manuel for JDemetra is not yet available but will be soon Meanwhile the User Manuel of Demetra can help out http www cros portal eu sites default files Demetra 20User 20Manual pdf Eurostat offers a variety of courses on seasonal adjustment and on the use of JDemetra several times a year For the exact dates and registration please check out the following link http epp eurostat ec europa eu portal page portal pgp ess about ess estp In case of specific questions on JDemetra please contact the Eurostat helpdesk under estat methodology ec europa eu 51
Download Pdf Manuals
Related Search