Title: Seasonal adjustment of time series observed at mixed frequencies using singular value decomposition with wavelet thresholding
Speaker: Wei Lin（Associate Professor, School of international economics and trade, University of international business and Economics）
Description: In this paper, we propose a novel seasonal adjustment method that accommodates the time series observed at mixed frequencies and possessing possibly multiple abrupt changes in seasonality, under the assumption that the nonseasonal component is difference stationary. Through a generalized difference, we remove the stochastic trend of the mixed frequency time series. Meanwhile, we express the seasonal component in terms of a matrix with a low rank SVD structure. The right and left singular vectors correspond to the seasonal patterns and their time-varying amplitudes. To estimate the SVD structure of seasonality and thus recover the seasonal component, we propose an effective algorithm that applies the wavelet thresholding technique to the left singular vectors. Our proposed method not only accommodates the persistence feature of seasonality, but also allows for the existence of possibly multiple abrupt changes in seasonality. Using both simulated and real data, we find that (i) when the seasonality is moderate or strong our proposed method performs well and correctly detecting the underlying seasonality structure; and (ii) for single frequency time series, the performance of our proposed method compares well with those of the traditional X-12-ARIMA and SEATS methods, especially in the case when the seasonality is strong.
Time: Apr 27, 2022（Wednesday），14:00-16:00
Venue: Tencent meeting Room ID：909 188160
