Modifications of SEATS Diagnostics for Under- and Overestimation

• Activity Number: 73
• Type: Topic Contributed
• Date/Time: Monday, August 9, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economics Statistics Section
• Author(s): David F. Findley*+ and Kellie Wills and Tucker S. McElroy
• Companies: U.S. Census Bureau and U. S. Census Bureau and U.S. Census Bureau
• Keywords: TRAMO/SEATS ; residual seasonality ; underadustment ; overadjusment ; stability ; sliding spans

Abstract

We consider SEATS diagnostic to determine whether, for an estimated seasonal decomposition component, there is underestimation or overestimation, meaning inadequate or excessive suppression of frequency components near the targeted suppression frequencies, e.g. the seasonal frequencies in the case of seasonal adjustment. The diagnostic in SEATS depends on variance estimates that assume an infinitely long filter has been applied. For the technically simplest case, estimation of the irregular component, we show that this results in substantial bias toward indicating overestimation. We consider two much less biased modifications of the diagnostic that are calculated from time varying variances associated with the finite-length filters and an associated test for the statistical significance of any indicated misestimation. The results indicate that this approach cannot detect overestimation reliably and is somewhat compromised in the detection of underestimation, because of the asymmetry of the distributions of the diagnostics.