Testing seasonal adjustment using Demetra Satybekova G B

Testing seasonal adjustment using Demetra+ Satybekova G. B. National Statistical Committee, Kyrgyzstan

Check of the original time series Original time series are: Accurate Long, for the years 2000 -2011, in average annual prices of the year 2000 (2000=100) Of good quality and correspond to the international standards Consecutive October 2011

Graphs that show seasonality Periodogram Auto-Regressive Spectrum There is a seasonal factor in the original dynamic range October 2011 Chart TSChart

Describe choice of the approach and predictors Approach used TRAMO/SEATS Certain national holidays were selected Automatically was selected specification - RSA 5 October 2011

Models used Information on pre-treatment: October 2011

Graphs of the results Yes, seasonal component was lost in the irregularity October 2011

Check for the sliding seasonal factor How to check the ratio of the sliding seasonal factor? October 2011

Main diagnostic of quality Basic test – compared the annual total value of the original series and the series adjusted for seasonal variations. This value should be close to zero October 2011

Residual seasonal factor Yes, there is residual seasonal factor after the adjustment October 2011

Stability of the model Model is stable October 2011

Remains of the normal distribution They are random October 2011

Some problematic series (if there were any) ? October 2011

Evaluate the ability to publish the results Need to learn from users about their needs and to expand their knowledge and understanding of the seasonal adjustment We think that it is too early to publish, we need to have certainty in the quality of the seasonal adjustment results received October 2011

Conclusion We need to have training, during the next seminar on how to analyse results of the seasonal adjustment How to document and publish the results October 2011