Dagum developed in 1996 a nonlinear nonparametric estimator of the nonstationary mean (trend-cycle) of monthly seasonally adjusted time series characterized by several points of maxima and minima. Relative to the widely applied classical 13-term Henderson filter (H13), this estimator reduces significantly both the size of the revisions to most recent estimates and the number of false turning points (unwanted ripples) with the good property of identifying true turning points with very short time delays. The purpose of this study is to develop a linear approximation to the Dagum nonlinear filter. A linear approximation offers the advantage of preserving the relationship between original and adjusted data when dealing with seasonally adjusted aggregates, it is of easy application for it does not required complex software, and its statistical properties can be confronted with those of other potentially competitive linear filters. This new linear filter is compared to the classical H13 by means of the classical spectral analysis techniques and statistical measures of bias, variance, and mean square error based on the system of weights.
A linear nonstationary mean predictor for seasonally adjusted series
DAGUM, ESTELLE BEE;LUATI, ALESSANDRA
2004
Abstract
Dagum developed in 1996 a nonlinear nonparametric estimator of the nonstationary mean (trend-cycle) of monthly seasonally adjusted time series characterized by several points of maxima and minima. Relative to the widely applied classical 13-term Henderson filter (H13), this estimator reduces significantly both the size of the revisions to most recent estimates and the number of false turning points (unwanted ripples) with the good property of identifying true turning points with very short time delays. The purpose of this study is to develop a linear approximation to the Dagum nonlinear filter. A linear approximation offers the advantage of preserving the relationship between original and adjusted data when dealing with seasonally adjusted aggregates, it is of easy application for it does not required complex software, and its statistical properties can be confronted with those of other potentially competitive linear filters. This new linear filter is compared to the classical H13 by means of the classical spectral analysis techniques and statistical measures of bias, variance, and mean square error based on the system of weights.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
