Seasonal Adjustment User Group meeting
held at INE of Portugal, in Lisbon
on the 20 of October 2011 (10h00 to 17h00), Room 206
Minutes and to-do-list
Presidency: Dominique Ladiray (FR)
Secretariat: Dario Buono (ESTAT), Faiz Alsuhail (FI)
Logistics: Martine Peeters (ESTAT)
Web: Agenda and presentations available here
Attendees (24): Dominique Ladiray (FR), Anna Ciammola (IT), Jekaterina Borodina (LV), Nomeda Bratcikoviene (LT), Regina Soares (PT), Miguel Balbina (PT), Roxana Ilie (RO), Faiz Alsuhail (FI), Alpay Kocak (TR), Mark Boxall (ECB), Jean Palate (NB, BE), Joerg Meier (NB,DE), Michael Richter (NB,DE), Sylwia Grudkowska (NB, PL), Peter Steiner (SECO), Emmanuelle Guidetti (OECD), Pascal Jacques (ESTAT), Dario Buono (ESTAT), Pilar Rey del Castillo (ESTAT), Martine Peeters (ESTAT), Gian Luigi Mazzi (ESTAT),Rosa Ruggeri Cannata (ESTAT), Simo Pasi (ESTAT)
Apologies (11): Valentina Gocheva (BG), Patrick Quill (IE),Isidoros Kypridakis (CY), Joslyn Magro (MT), Daniel Ayoubkhani (UK), Marcus Scheiblecker (WIFO), Jukka Jalava (ESTAT),Hervé Rennié (ESTAT), Hannah Kiiver (ESTAT), Maria del Carmen Gonzalez (ESTAT), Karo Nuortila (ESTAT), Roberto Barcellan (ESTAT)
Document history:
Version / Release Date / Author / Reviewed By1.0 / 04/11/2011 / Faiz Alsuhail (FI) / Dominique Ladiray (FR)
Dario Buono (ESTAT
Table of Content
(1) Approval of the agenda, the minutes and to do -list of the previous meeting 2
(2) The next release of J-Demetra 2
(3) Explanation and interpretation of the Tramo/Seats results in Demetra+ 4
(4) Using Demetra+ in daily work 4
(5) Reflection on the Benchmarking methodology to be included in J-Demetra 5
(6) Validation of the User Manual 5
(7) The coming ESTP-courses on Demetra+ and seasonal adjustment 5
(8) Specification of TD effects for Demetra+ 6
(9) To-do-list: Contribution on SAUG members to additional modules to be developed in J-Demetra 6
(10) Pending questions sent by the User Members to the ESTAT methodology help desk 7
(11) Any other business 7
(12) Next meeting 8
(1) Approval of the agenda, the minutes and to do -list of the previous meeting
The chair of the meeting, Mr. Dominique Ladiray, opened the meeting and welcomed everyone to Lisbon.
The minutes were approved. As for the list of tasks to be completed, it was noticed that:
- A FAQ on “How to deal with a bunch of series when some of them do not have to be seasonally adjusted” has been answered and is available at CROS.
- A note on the Excel add-ins has been included in the User Manual.
- A note on the WSA cruncher has not been done (INSEE is in charge of this work).
- The first work on benchmarking modules has been completed.
- A note on direct vs. indirect seasonal adjustment with Demetra+ has been written and an abstract of the note has been sent to CFE11 to take place in London in December 2011.
- The specification of a calendar module has not been done but a presentation on the main points will be done during this meeting.
(2) The next release of J-Demetra
Overview
Mr. Jean Palate gave a short presentation where he talked about the latest developments on the .net-version on Demetra+ (hereafter Demetra+). A more detailed discussion can be found from his (power point) presentations. He also talked about the Java-version of Demetra+ (hereafter J-Demetra). A more detailed discussion can be found from his PDF-presentation.
There will be a new build of Demetra+, which will be released on 30th October. This version will be corrected for bugs and include slight improvements in the user interface.
A new release, Demetra+ 1.1 will be out by 31st December. This release will include modifications and improvements in the core engines and multiprocessing. Most of the improvements are related to Tramo/Seats; only a few of the changes concern X12. One of the important modifications (in the Tramo/Seats-part) concerns the automatic model identification.
The J-Demetra is under development. In this new software it is possible to use separate modules (and not necessary the whole software). Hence, the software should be easy to plug in to a production chain. New features can be programmed/ introduced by using the core modules.
Discussion
Some users felt, that making changes in the automatic model identification is not good. It was felt, that there needs to be a certain amount of consistency and stability in the methods. Too many changes in the algorithms lead to revisions in the (SA) results, which is not wanted.
Others felt that the latest changes in the Tramo/Seats algorithms must be taken into account in the software development. They underlined, that changes in the automatic model identification doesn’t introduce such big changes since models are kept unchanged (and only parameters are re-estimated) in the month-to-month and quarter-to-quarter production. When models are changed the (SA) results will change as well, no matter what kind of model identification procedures one is using.
It was decided that the bugs in Demetra+ will be corrected. However, the version with new automatic model identification will not be distributed outside the user group. The bug-corrected version of Demetra+ will be used at the coming ESTP-course in November.
When it comes to J-Demetra, the members felt that the features that are being developed should be coherent with the guidelines.
At the end, the group felt there was a need for a clear release policy for the software.
(3) Explanation and interpretation of the Tramo/Seats results in Demetra+
Overview
Alpay Kocak presented his paper, where he explains the (diagnostic) results of Tramo/Seats in Demetra+. He covered the main results, results from the pre-processing (Tramo) and explained things related to the decomposition (Seats).
A more detailed discussion can be found from his power point presentation or paper.
Discussion
Ms. Sylwia Grudkowska informed that the description of diagnostic results of Tramo/Seats in Demetra+ User Manual is based on Mr. Alpay Kocak paper, significantly extended by additional information from Mr. Agustin Maravall papers and supplemented by explanations provided by Mr. Agustin Maravall.
The discussion was mostly around the documentation around SA and Demetra+. Some felt that there should be a book explaining “SA for dummies”. The Handbook on Seasonal Adjustment, which is supposed to be ready in 2012 (the review process should begin very soon), isn’t going to be such a book.
Mr. Mazzi took the opportunity to inform, that next year (2012) there will be a task force which will revise and possibly update the ESS Guidelines on Seasonal Adjustment.
(4) Using Demetra+ in daily work
Overview and Discussion
INE presented results and shared experiences on using J-Demetra in daily production. A detailed discussion can be found from their power point presentation.
Three problems were brought up in the presentation and discussed. The first was about Easter regressors. It was noticed that the value of the regressor is different when one uses Tramo/Seats and X12. The reason behind this is in the different way regressors are normalized in these two methods.
The other point, which was discussed, is about terminology. J-Demetra computes a series, which is called Seasonally Adjusted series, though it doesn’t include the effect of outliers. The outlier effect is included in the Final Seasonally Adjusted series. This was found confusing.
The last point was about the content and labeling of the output. Some series had acronym that did not help too much in telling, what the specific component was.
(5) Reflection on the Benchmarking methodology to be included in J-Demetra
Overview
Mrs. Bratcikoviene gave an overview of the existing benchmarking methods and suggested some diagnostics to be checked after benchmarking.
Discussion
It was agreed, that direct and indirect and benchmarking are two separate problems. Making constrains for the series to be benchmarked was not seen as an easy task. J-Demetra will however have an option for (univariate) benchmarking. It was seen, that benchmarking should be a separate module in J-Demetra.
It was agreed to add some benchmarking tool to Demetra+ and J-Demetra. When it comes to diagnostic tool the benchmarking SAS-procedure of Statistics Canada will be consulted for inspiration.
(6) Validation of the User Manual
Overview
Mrs. Grudkowska presented the latest changes in the User Manual and suggested, that the document would be read once again before being published.
Discussion
Some members felt that the manual emphasizes Tramo/Seats too much in comparison to X12 and would hence be biased. Mr. Ladiray promised that he would write a few extra pages about X12 to make the presentation more balanced. The SA Steering Group will finally approve the manual.
Some felt that the manual is too long and includes too much information about the methods used. Others, however, felt that the scope and content of the document was not too wide, especially in comparison with other manuals (SAS, STATA, E-Views etc).
UNECE has published a user guide on Demetra+. This document, however, presents only seasonal adjustment by Tramo/Seats. It was agreed, that this document will not be circulated through CROS.
(7) The coming ESTP-courses on Demetra+ and seasonal adjustment
Overview
There will be an advanced ESTP course on seasonal adjustment and Demetra+ in Luxemburg on 15th to 17th of November. Mr. Buono received 34 applications to the course. There was room for just 14 participants.
It was underlined that only one delegate per country was allowed to participate. The selected students were most in line with the target group of the course, that is, were advanced users of seasonal adjustment methods and capable of spreading the knowledge in their home institutions.
Discussion
There will be three ESTP courses on seasonal adjustment and Demetra+ on 2012. The first one will take place on January 24- 26 "Introduction to Seasonal adjustment and Demetra+"and is meant to beginners. In June 26-28 there will be "Advanced Seasonal Adjustment" and is meant for people willing to become experts. In November 13-15 "Demetra+ for advanced users" and is meant to people who are already experts.
It was asked whether it was possible to increase the number of participants in one course. Many people are carrying laptops and hence the size of a computer lab should not be a reason to leave some people out of a course. On the other hand, it was felt that a large number of students is difficult to manage with the given teacher resources. Therefore a small course would be more ideal than a large one.
(8) Specification of TD effects for Demetra+
Overview and discussion
Mr. Ladiray gave a presentation on trading day effects. His presentation was mostly about the construction of a set of regressors taking into account National calendars and the measurement of Ramadan effect (a specific moving holiday effect). A more detailed discussion can be found from his presentation.
(9) To-do-list: Contribution on SAUG members to additional modules to be developed in J-Demetra
Overview and discussion
The following to do -list was agreed on:
What? / Who? / When?Take a leading role when it comes to the documentation and release policy of Demetra+ and J-Demetra / Eurostat, ECB / In the future.
Give a proposal for quality check. / Dominique Ladiray, Sylwia Grudkowska and Anna Cimmola / The proposal will be given in the next meeting of the user group.
Give a proposal on how to make the CSV output more user-friendly / INE (Portugal) / The proposal will be given in the next meeting of the user group.
Write a few pages on X12 and the text will be added to the User Manual of Demetra+. Then the manual will be sent to SASG for approval. / Dominique Ladiray / November 20th
Install the Demetra+ corrected for bugs on the Eurostat PCs so that can be used in the coming November 2011 ESTP course. The manual will not be inside the software. / Dario Buono / November 2011
Before the User Manual is approved by the SASG there will be a disclaimer in the User Manual / Mark Boxal
write a proposal on how to do univariate benchmarking on Demetra+ / Dominique Ladiray, Nomeda / The proposal will be given in the next meeting of the user group.
make a proposal to the SA steering group taking place on the 21 of November 2011 for the content of the ESTP courses of 2012 and try to find out what are the available trainer resources / Dario Buono / Before 21 of November.
(10) Pending questions sent by the User Members to the ESTAT methodology help desk
Overview
Buono told that he had replied to some of the questions sent to the Eurostat helpdesk. Some questions were forwarded to other experts: Kocak, Grudkowska and Palate.
Discussion
There was a question of whether the helpdesk should reply to questions, which are sent by individuals working at private institutions or somewhere outside NSI’s. The reaction was that the helpdesk should serve everyone, as it is a service financed by European taxpayers.
(11) Any other business
There is a new one-year contract with the National Bank of Belgium about developing J-Demetra. It was hoped, that testing the new tool should start soon.
INSEE is having SA days on 19th-21st of December. Mr. Ladiray invited everyone to listen and to give presentations. The language of the seasonal adjustment days is French. Participants should cover their travel expenses.
There was an idea of having a Demetra+ course for programmers. The course would be about modifying the modules and making own applications. The future of such a course was left open.
(12) Next meeting
The idea is to have three meetings in 2012: one in February, another in June and the third one in October.
Countries willing and able to host a meeting by providing a meeting room for 20-30 people and offering coffee during the breaks are requested to let mr. Buono know.
4