CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Alternative methods of seasonal adjustment Authors:  Stephen Pollock - University of Leicester (United Kingdom) [presenting]
Abstract: The conventional methods of seasonal adjustment rely on filters realized in the time domain that nullify the sinusoidal elements of a data sequence that are to be found at the seasonal frequencies at its harmonics. Such filters combine a low pass filter that attenuates the high frequency elements of the data with a so-called comb filter that eliminates the seasonal frequency and its harmonics. Typically, the filters are derived from the estimates of a structural time-series model or of a reduced-form ARMA model. Often such filters fail fully to nullify data components that are adjacent to the seasonal frequencies, which serve to modulate the patterns of seasonal variation. They may leave a residue of the seasonal fluctuations in the filtered data. The aim is to analyze the effects of the common methods of the seasonal adjustment and to propose alternative methods that operate in the frequency domain and that enable a careful choice to be made of the seasonal elements that should be eliminated from the data.