Ketty Attal-Toubert and Stéphanie Himpens Insee, France 16th of November, 2011 ESTP course Demetra+ Demetra+ for X12 in Daily Work.

Slides:



Advertisements
Similar presentations
Testing Relational Database
Advertisements

Process Monitoring is only the first step in improving process efficiency.
United Nations Statistics Division/DESA
MASTER QUOTE OVERVIEW.
Exercise 7.5 (p. 343) Consider the hotel occupancy data in Table 6.4 of Chapter 6 (p. 297)
Decision Analysis Tools in Excel
Test Case Management and Results Tracking System October 2008 D E L I V E R I N G Q U A L I T Y (Short Version)
Seasonal Adjustment of National Index Data at International Level
TRACK 2™ Version 5 The ultimate process management software.
Internal documentation and user documentation
REPUBLIC OF TURKEY TURKISH STATISTICAL INSTITUTE TurkStat Direct vs. Indirect Approach in Seasonal Adjustment: Proposal for a new tool Necmettin Alpay.
Welcome to Turnitin.com’s Peer Review! This tour will take you through the basics of Turnitin.com’s Peer Review. The goal of this tour is to give you.
X-12 ARIMA Eurostat, Luxembourg Seasonal Adjustment.
TRACK 3™ The ultimate process management software.
Seasonal Adjustment Methods and Country Practices Based on the: Hungarian Central Statistical Office: Seasonal Adjustment Methods and Practices; UNECE.
The Project AH Computing. Functional Requirements  What the product must do!  Examples attractive welcome screen all options available as clickable.
Tutorial 5: Working with Excel Tables, PivotTables, and PivotCharts
J. Khélif Insee July 2008 A quality report for seasonally and trading day adjusted French IIP.
Demetra+ Quick Tour Versatile software. Choose the right tool Demetra+ main feature: multi-processing Demetra+ in production. Understanding.
Term 2, 2011 Week 1. CONTENTS Types and purposes of graphic representations Spreadsheet software – Producing graphs from numerical data Mathematical functions.
Report Manager for FSGs in R12 Farewell ADI toolbar
Moodle (Course Management Systems). Assignments 1 Assignments are a refreshingly simple method for collecting student work. They are a simple and flexible.
Ts_print in a few easy steps There are four screens: Entities, Data Items, Date, and Report Format.
Copyright © 2007, Oracle. All rights reserved. Managing Concurrent Requests.
USING DEMETRA+ IN DAILY WORK SAUG – Luxembourg, 16 October 2012 Enrico INFANTE, Eurostat Unit B1: Quality, Methodology and Research.
United Nations Economic Commission for Europe Statistical Division Seasonal Adjustment Process with Demetra+ Anu Peltola Economic Statistics Section, UNECE.
Rev. 0 CONFIDENTIAL Mod.19 02/00 Rev.2 Mobile Terminals S.p.A. Trieste Author: M.Fragiacomo, D.Protti, M.Torelli 31 Project Idea Feasibility.
Publishing Seasonally Adjusted Data Anu Peltola Economic Statistics Section, UNECE UNECE Workshop on Seasonal Adjustment 20 – 23 February 2012, Ankara,
Overview of Main Quality Diagnostics Anu Peltola Economic Statistics Section, UNECE UNECE Workshop on Seasonal Adjustment 20 – 23 February 2012, Ankara,
Reports and Learning Resources Module 5 1. SLMS Primary Administrator Training Module 5: Reports and Learning Resources 2.
1 Departamento de Contas Nacionais / Serviço de Indicadores de Curto Prazo National Accounts Department / Short Term Statistics Unit Using Demetra+
Testing Seasonal Adjustment with Demetra+ Levan Gogoberishvili Head of National Accounts Division National Statistics Office of Georgia.
Rev.04/2015© 2015 PLEASE NOTE: The Application Review Module (ARM) is a system that is designed as a shared service and is maintained by the Grants Centers.
XP. Objectives Sort data and filter data Summarize an Excel table Insert subtotals into a range of data Outline buttons to show or hide details Create.
Ketty Attal-Toubert and Stéphanie Himpens Insee 22nd of June, 2011 An Overview of seasonal adjustment in the short term statistic department.
Testing seasonal adjustment with Demetra+ Dovnar Olga Alexandrovna The National Statistical Committee, Republic of Belarus.
McGraw-Hill/Irwin The Interactive Computing Series © 2002 The McGraw-Hill Companies, Inc. All rights reserved. Microsoft Excel 2002 Working with Data Lists.
April 2011 Testing Seasonal Adjustment with Demetra+ Ariunbold Shagdar National Statistical Office, Mongolia.
Harmonisation of Seasonal Adjustment Methods in EU and OECD Countries Ronny Nilsson Statistics Directorate.
Excel part 5 Working with Excel Tables, PivotTables, and PivotCharts.
IMF Statistics Department The views expressed herein are those of the author and should not necessarily be attributed to the IMF, its Executive Board,
Economic Data Sources on the Internet Su, Chapter 3.
Using Demetra+ at SURS Manca Golmajer, Andrejka Smukavec 20 June 2013.
HRP Copyright © Leland Stanford Junior University. All rights reserved. Warning: This presentation is protected by copyright law and.
USING DEMETRA+ IN DAILY WORK SAUG – Luxembourg, 16 October 2012 Enrico INFANTE, Eurostat Unit B1: Quality, Methodology and Research.
Testing Seasonal Adjustment of the Price Index for tomatoes with Demetra+ Kumpeisova Dinara Agency of Statistics of the Republic of Kazakhstan, Kazakhstan.
Aggregator Stage : Definition : Aggregator classifies data rows from a single input link into groups and calculates totals or other aggregate functions.
Ketty Attal-Toubert and Stéphanie Himpens Insee 22nd of June, 2011 Using SAS to implement additional tools.
1 ConIstat-on-line Short term statistical time-series data base in web environment By Francesco Rizzo
FAST STUDENT Your Chance to Learn!. FAST STUDENT Your Chance to Learn!
Tutorial 5: Working with Excel Tables, PivotTables, and PivotCharts
Data Virtualization Tutorial: XSLT and Streaming Transformations
Shohreh Mirzaei Yeganeh United Nations Industrial Development
Using Demetra+ at NSO of FI, - first results of testing
Built by Schools for Schools
Quality Reporting with JD+
WinTIM, Indices methodology and tool Wiking Althoff, CESD Communautaire External trade experts meeting on the CARDS Programme, Luxembourg, May.
How to select regressors and specifications in Demetra+?
New Demetra 2.2 Euro-indicators Working Group
STATISTICAL AGENCY UNDER PRESIDENT OF THE REPUBLIC OF TAJIKISTAN
Tutorial 7 – Integrating Access With the Web and With Other Programs
Faiz Alsuhail 21 of June, Frankfurt
Demetra+ User Manual Final release
Tools to make the process safer : secured filing
TYPES of seasonal adjustment processes in banco de españa
Model Selection, Seasonal Adjustment, Analyzing Results
Ermurachi Galina National Bureau of Statistics, Republic of Moldova
Assignment resource Working with Excel Tables, PivotTables, and Pivot Charts Fairhurst pp The commands on these slides work with the Week 2 Excel.
Launching of an EAP campaign (EAP =Enquête annuelle de production => annual survey on industry) EUROSTAT – 15/16 october 2013.
Using JDemetra+ at STATEC
Presentation transcript:

Ketty Attal-Toubert and Stéphanie Himpens Insee, France 16th of November, 2011 ESTP course Demetra+ Demetra+ for X12 in Daily Work

Page 2 Contents of the presentation › 1. Overview of seasonal adjustment at Insee’s short term statistics department › 2. Which tools are used? › 3. Using SAS to implement additional tools › 4. Using Demetra+ to improve quality › 5. How to make the process safer: controlling refreshment › 6. How to make the process safer: filing

Page 3 An overview of seasonal adjustment at Insee’s short term statistics department

Page 4 Which Series are seasonally adjusted ? › IPI : 91 series, NACE3 Level › Turnover Indexes : 440 series, NACE (3-digit) in industry and NACE (5-digit) in other sectors. › All are monthly series.

Page 5 The seasonal adjustment process › According to ESS Guidelines on seasonal adjustment, we implement two processes : – Each year a complete review of the models – Each month a re-estimation of the parameters of the models defined by the annual process. Outliers are re- identified on the last period. => Partial concurrent adjustment

Page 6 The seasonal adjustment process › Indirect seasonal adjustment : Seasonally and calendar adjusted series of the aggregates are computed as a weighted sum of their constitutive seasonally and calendar adjusted series.

Page 7 Methods and tools › Since 2011 we use X12-ARIMA in Demetra+. › We developed a set of SAS macros to do some steps of the annual and monthly campaigns.

Page 8 The annual seasonal adjustment campaign The annual campaign is done in 8 stages : 1.Specification of the relevant set of Trading day regressors. Different sets of regressors are considered depending on the sector of activity (retail trade, wholesale trade, services, industry); 2.Automatic identification of ARIMA model and outliers (X12- Demetra+), with selected TD-regressors. Quality report on new and old treatments are computed; 3.Comparison of quality reports is made in order to automatically select the best ARIMA model for each series; 4.Interactive check of results and search for best specifications for problematic series in order to improve quality; 5.New computation of quality report, if interactive changes of specifications; 6.Graphical comparison between new and old seasonally adjusted series and return to interactive analysis if needed; 7.Quality report of the treatment at aggregate level (new and old). 8.Final search for better specifications for series which cause a decline in quality of aggregates.

Page 9 Annual SA campaign Specification of relevant set of regressors Automoatic identication of model (X12 - Demetra+) Interactive search of new specifications. Demetra+ shows poor diagnostics Quality reportGraphical comparison new/old SA series Quality report for aggregates (old and new treatments) Unsatisfying quality report Comparison of quality reports => Automatic selection of the best models for each series Unsatisfying revisions Unsatisfying quality report

Page 10 Monthly SA campaign 1.Update of the raw data file 2.Seasonal adjustment in Demetra+, launched from SAS. 3.Interactive check in Demetra+ of seasonal adjustments results and change of some specifications if needed. 4.Export of results 5.Saving of the workspace, the linked data file and the specifications in an archive directory.

Page 11 Which tools are used ?

Page 12 Demetra+ › To seasonally adjust the series (but execution is ordered from SAS (WSACruncher)). › To check quality : detailed examination of results and of diagnostics › To modify specifications in an interactive way if quality is not good enough

Page 13 Launching Demetra+ from SAS (WSACruncher) › It allows to launch seasonal adjustment from SAS which is the tool in which most of other treatments are done. Allows to generate output (result series and matrix view). Results can then be imported in SAS format as soon as seasonal adjustment is finished. Example : wsacruncher d:\demetra+\test_1.xml -t -p complete › => It limits changes of environment in the computing process of indices, which increases safety of the process. Interactive Demetra+ is only opened to check quality and results.

Page 14 SAS › To implement additional tools › To increase safety and traceability of the processes => More details later

Page 15 Excel › Input and output of Demetra+: –Raw data in SAS format are converted to Excel format for being used in Demetra+ –Output generated by Demetra+ (output series + matrix view in.csv format) are converted to SAS format. › More attractive presentation of quality report results : SAS tables from quality report are converted to Excel format. Colours are used to spot problematic series.

Page 16 Tools used in annual process: summary Sas Building and selection of trading day regressors Creation of specifications sets for automated treatment Export of raw data to an Excel file Launching of seasonal adjustment treatment and generation of output with cruncher. Demetra+ Execution of seasonal adjustment treatment Detailed analysis of the treatment Interactive change of specifications. New generation of output Sas Import of results from Excel to SAS format Quality report Comparison of quality reports from different treatments Saving of final annual process results (workspace, data file, specifications) Excel Raw data file Excel Output of Demetra+ Excel Results of quality report

Page 17 Tools used in monthly process: summary Sas Updating of raw data file Export of raw data to an Excel file Launching of seasonal adjustment refreshment and generation of output with cruncher Demetra+ Refreshment of seasonal adjustment Detailed analisys of the treatment Interactive change of specifications if necessary. New generation of output Sas Import of results from Excel to SAS format Saving of monthly process results (workspace, data file, specifications) Tool to compare specifications betwen month m and month (m+1) Excel Raw data file Excel Output of Demetra+

Page 18 Using SAS to implement additional tools

Page 19 Summary of the cases in which SAS is used to implement additional tools › To identify and compute sets of regressors › To produce a detailed quality report › To compare quality reports, graphs and results of seasonal adjustments processed using different sets of specifications › To automatically change specifications when they lead to poor results › To immediately see what has been changed in the specifications of a workspace between month M and M+1

Page 20 Computing and identifying sets of trading-day regressors (1) › We developed a specific way to compute our calendar regressors which take into account both French calendar and specificity of each sector of activity under review. › In TRamo-Seats and X12-ARIMA, predefined regressors consider only two cases: Trading-day effect (6 variables: all days are different) or Weekday effect (Monday until Friday are similar but different from Weekend days). › Those cases are not suitable for all sectors of activity, like French retail trade (low activity on Monday, high activity on Saturday)

Page 21 Computing and identifying sets of trading- day regressors (2) › Automatic selection is made among several sets of regressors (more choice than pre-defined regressors). Criteria for selection: combination of Fisher test (covariance matrix of coefficients in given in Dos version of X12-ARIMA), and Akaike Criterion (AICC, given in Demetra+). › Available sets depend on the sector of activity of the series. Sector of activityNumber of available sets of TD-regressors Industry3 Retail and Wholesale trade, Services 4

Page 22 Computing and identifying sets of trady-day regressors (3) Output of the SAS program :

Page 23 Checking quality (1): an additional tool Aims of this quality report, built in 2007: › Select criteria among many and add some complementary diagnostics › Synthesize this information for each series: – control of the way the quality criteria are summarized – makes the comparison of the quality of different treatments easier › Synthesize information at aggregate level: – way of judging the quality of aggregates – allows a top-down analysis of the quality and to target those series that are “really” problematic Solution: › Assign grades to evaluate different aspects of quality; › Synthesize them for each series (weighted mean of the grades, the weight depending on the importance of the quality criterion); › Aggregate the grades of the series according to the weight of each series in the aggregate series (economic weight).

Page 24 Checking quality (2): computing grades › Grades (numerical values) are computed by SAS programs. In the columns below the grade is 100 if the null hypothesis of the test is not rejected at 5% level, 50 if it is rejected at 5% but not rejected at 1% level, 0 if it is rejected at 1% level. The overall quality of ARIMA model is computed as a weighted mean of the grades related to the quality of residuals

Page 25 Checking quality (3): groups of diagnostics The different groups of diagnostics are : › ARIMA modelling quality › Outlier identification (number of outliers, concentration) › Quality of decomposition (residual seasonality or residual trading-day effects, orthogonality of components) › History revisions (+outliers stability) Weighted means of grades are computed to evaluate the quality of the aggregates (indirect seasonal adjustment). Results are summed up in an Excel file.

Page 26 Comparing quality and results › SAS output of the quality report is exported in Excel file. To compare the quality of two different treatments for a given series, the difference between the synthetic grades are calculated => allow to spot series which show a decline or an improvement of quality. › The quality report also produces a file which contains the characteristics of the treatment (ARIMA model, transformation outliers..). It is then possible to spot differences between the two treatments. › Graphs overlaying results of different treatments are drawn in SAS. It is possible to easily scroll the results.

Page 27 Graphical comparison of old and new SA series (SAS)

Page 28 › It is easy to read and write specifications in SAS (easy link with XML files). Changing specifications › We create a SAS table containing specifications Data file specifications Arima model parameters Series span specifications

Page 29 Changing specifications automatically SAS allows us › to create mix models (for example old ARIMA models with new regressors). › To fix parts of models (the selected set of regressors for each series). › To change specifications in case of bad diagnostics (to test a model with the maximum number of calendar regressors every time residual calendar effects have been detected).

Page 30 Comparing two sets of specifications (1) Are the two sets of specifications different or not ? Do they use the same data file ? Do they share the same options to find ARIMA models (initial specifications) ?

Page 31 Comparing two sets of specifications (2) Two types of specifications can be compared : › It is possible to compare results : the models used to process the series with no care of whether they’re a result of an automated identification or if they were fixed by the user › It is also possible to compare initial options : which process leads to this model ? => Traceability of the history of changes in specifications from month to month.

Page 32 Using Demetra+ to improve quality

Page 33 Interactive use of Demetra (1) › To do a quick check of all series during annual and monthly campaigns: Mains Results menu and Charts menu.

Page 34 Interactive use of Demetra (2) › For a detailed analysis of problematic series: – « Bad» or « Severe » according to Demetra+ – New automatically identified transformation of the series (logarithm or none) different from the old one. – Charts looking « strange » or very different from the old ones. – Degradation of the quality report grades. Priority is given to series with highest weight in the global index.

Page 35 How to make the process safer: controlling refreshment

Page 36 Refreshment › Monthly process : –A new value is added at the end of the series in the data linked file and, generally, previous raw data are revised. –Seasonal adjustment is done in Demetra+. => To take changes in the raw data file into account, refreshment must be performed on the SAProcessing.

Page 37 2 kinds of specifications › An initial set of specifications (for example automatic ARIMA model identification, test for transformation…) › A final set of specifications (what has been found for each part of the model : ARIMA model, transformation etc.) = specifications from result

Page 38 Refreshing options There are different options of refreshment (Concurrent, partial concurrent, current) but in all cases: › They read the linked raw data file to update the data contained in the processing. › They take into account initial specifications.

Page 39 Refreshing options › Example : –Suppose we decide to do a full automated process except for outlier identification, which is disabled. –A « Partial concurrent, all outliers » refreshment is then done. One might think outlier identification would be performed on the full span of data since outliers are supposed to be refreshed…

Page 40 Refreshing options No new identification is done ! => No outlier identification was requested in initial specifications.

Page 41 Refreshing options A few remarks : › In a refreshment process, treatments applied to each series can be different according to the differences in their initial specifications › There is no way to easily check what is done for each series

Page 42 Refreshing options What we want (in monthly campaigns): › The models to be fixed (ARIMA part, outliers before a fixed date, calendar regressors, transformation) › The outliers to be identified only from a fixed date (Demetra+ doesn’t allow to do this). => SAS

Page 43 Refreshing options A new SAProcessing XML file is written in SAS in which: › Every items of the specifications are fixed. Result specifications of the annual campaign are turned into initial specifications in the monthly processing, but: – Outliers which occur after a fixed date (ex. earliest date of the raw data revision period) are removed (except for some of them we want to keep); – Outlier identification is requested from this fixed date. › “Concurrent” refresh option is then selected (=“complete” in cruncher statement).

Page 44 How to make the process safer: filing

Page 45 Why should the filing of a workspace be made safer ? (1) What we copy :

Page 46 Why should the storing of a workspace be made safer ? (2) › Data saved in the processing : –The specifications –The path of the linked data file › The path to the user-variables is recorded in the Variables xml file › Every change in the original file might lead to a change in the filed workspace. Linked to the original data file and to the user- variable data file

Page 47 What must we do in a filing process to secure our results ? › To secure a workspace: –The workspace of the monthly or annual process must be copied in an appropriate directory. –The path of the linked data files must be changed into the ones copied in the archive directory.

Page 48 The different steps › Copy of the workspace in the archive directory › Copy of the data ending in month M and of the user- variables file in the archive directory › In the archive directory: – change of the processing so that it points towards the file saved in the archive directory –change of the user-variable XML file so that it points towards the file saved in the archive directory › Additionally, initial and result specifications read from SaProcessing XML file are stored in SAS tables. › All steps are done in SAS.

Page 49 Demetra+ workspaces for seasonal adjustment at Insee’s short term statistics department (1) Suppression of outliers from a chosen date (except some of them we want to keep) Workspace of monthly process Initial specifications Result specifications Refresh « complete » with cruncher Archived workspace (month m) Filing process Month m Final workspace of annual campaign Initial specifications Critical value for outliers detection Result specifications Archived workspace (year y) Filing process Year y

Page 50 Demetra+ workspaces for seasonal adjustment at Insee’s short term statistics department (2) Result specifications Workspace of monthly process Initial specifications Refresh « complete » with cruncher Archived workspace (month m) Filing process Month m Workspace of monthly process Initial specifications + possible interactive changes Result specifications Refresh « complete » with cruncher Archived workspace (month m+1) Filing process Month m+1 Updated raw data