Filtering in the Fourier domain

19/06/2017 15:34
The design and use of a new linear filter to adjust time series for seasonal effects
A linear filter is proposed in this paper, which can be incorporated in existing frameworks and packages for the treatment of time series, and which identifies seasonal influences in one step, with no iteration required. The aim of the study is to evaluate the efficacy of this proposed filter. In particular, the filtered series is used as the input of a seasonal adjustment procedure, performed by JDemetra+. The performance of the proposed filter is analyzed on a real time series;
most of the observed diagnostics focus on spectral analysis.