Evaluation of Finite-Sample Diagnostics for Model-Based Seasonal Adjustments and Trends

• Activity Number: 52
• Type: Topic Contributed
• Date/Time: Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
• Sponsor: Business and Economics Statistics Section
• Author(s): David Findley*+ and Richard Gagnon and Tucker S. McElroy
• Companies: U.S. Census Bureau and U.S. Census Bureau and U.S. Census Bureau
• Keywords: ARIMA ; signal extraction ; SEATS ; X-13A-S

Abstract

Most seasonal adjustment and trend diagnostics of the SEATS seasonal adjustment program (Gomez and Maravall, 1997, Maravall and Caporello, 2004) are calculated with formulas for infinite-length series. Using matrix formulas of Bell and Hillmer (1988), McElroy and Sutcliffe (2004), and McElroy (2005) for finite-sample ARIMA model-based signal extraction filters and the error covariances of their estimates, a variety of finite-sample-based diagnostics were obtained. Findley and Martin (2003, 2006) and Findley, McElroy, and Wills (2004) show early examples. Now, finite-sample alternatives to the main diagnostics of SEATS and new diagnostics have been developed and implemented in a test version, the U.S. Census Bureau's hybrid X-13A-S program fusing X-12-ARIMA and SEATS. This has enabled an evaluation study of these diagnostics involving many series, results from which we will summarize.