A Study of Diagnostics for Detecting Seasonality and Residual Seasonality

• Activity Number: 593
• Type: Topic Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): David Findley*+ and Demetra Lytras
• Companies: U.S. Census Bureau and U.S. Census Bureau
• Keywords: Time series ; Seasonal adjustment ; Moving seasonality ; Sasonal autoregressive models

Abstract

Using an empirical study with Census Bureau time series and moving seasonality simulations from first order seasonal autoregressive models, we evaluate and compare the properties of four diagnostics for detecting seasonality and residual seasonality. These are the autoregressive spectrum diagnostic's implantation by Soukup and Findley (1999), the Tukey spectrum test of Maravall (2012), the QS (quasi-) chi-square statistic of Maravall (2012) for positive seasonal autocorrelation, and the Generalized Least Squares regression F-test for stable seasonality of Lytras, Feldpausch and Bell (2007). All are available in the latest versions of X-13ARIMA-SEATS and TRAMO-SEATS. The properties of these diagnostics have not been adequately investigated, especially regarding detection of residual seasonality, for which we show that shorter data spans are required for good performance with three of the diagnostics.