The U.S. Census Bureau is in the process of incorporating an additional ARIMA (Auto-regressive Integrated Moving Average) model selection procedure into the X-12-ARIMA software. The automatic modeling procedure in X-12-ARIMA Version 0.2 examines a list of five possible ARIMA models (found in the file x12a.mdl). The new procedure in Version 0.3 has a broader range of models to choose from, and we expect it to fit models to a wider range of series than Version 0.2 (Monsell, 2002). In addition, Version 0.3 has more options than in previous versions.
In this paper, we compare the two automatic modeling procedures. To compare the procedures, we modeled a large group of Census Bureau series including U.S. Imports, Exports, and Retail Sales series. We compared models using diagnostics such as the spectral peaks, Ljung-Box Q (LBQ) statistics, and forecast errors. The goal of our research is to not only determine which procedure performs better, but also give analysts some guidance on the various options available in the new procedure.
Our motivation behind this study was two-fold: 1) to provide users with a documented study of the benefits of the new method, and 2) to give the users some ideas of what to expect from the new automatic modeling procedure once the new version is released.
seasonal adjustment, time series
(3) Kathleen M. McDonald-Johnson is Mathematical Statistican, Economic Statistical Methods Division, U. S. Census Bureau, 4600 Silver Hill Road, Washington, DC 20233. email : firstname.lastname@example.org
Roxanne M. Feldpausch is currently a consultant.
This symbol indicates a link to a non-government web site. Our linking to these sites does not constitute an endorsement of any products, services or the information found on them. Once you link to another site you are subject to the policies of the new site.