【主讲】林蔚 (首都经贸大学)
【主题】Flexible Seasonal Adjustment Methods for Time Series Data Using Regularized Singular Value Decomposition
【时间】2016年4月8日 (周五) 15：30-17：00
【地点】上海财经大学经济学院楼701室
【语言】英文
【摘要】In this paper, we propose a new seasonal adjustment method based on regularized singular value decomposition (SVD) of the matrix which reshapes the seasonal time series data. In the regularized SVD, the right singular vectors capture multiple seasonal patterns, and the corresponding left singular vector captures the magnitudes of those seasonal patterns and how they changes over time. The proposed method applies to seasonal time series data with deterministic seasonal component and stationary/nonstationary nonseasonal component: We impose roughness penalties on the left singular vectors to identify the deterministic seasonal component with the magnitudes of seasonal patterns change smoothly with and without breaks. The new method, fully data-driven and flexible, outperforms X13 seasonal adjustment method by delivering smaller losses for stationary time series with salient seasonal component, and for differenced stationary time series in simulation exercises.
