L’effet des points aberrants dans la désaisonnalisation The presence of outliers in the estimation of the seasonal factors

ABSTRACT

This article is devoted to an improved estimation of the seasonal factors in the X-11-ARIMA method given the presence of outliers in the unadjusted series. The series are modelled by an ARIMA process and the outliers are identified relative to the fitted values of the model. They are then replaced by their corresponding function values. A good replacement of the outliers improves both the performance of the ARIMA model fitted to the modified series and the revisions to the seasonal factors, especially the latter.