【主讲】周雨田 教授 (台湾中央研究院经济研究所)
【主题】A New Normalization Method for Studying Time-Frequency Variations of Data with High Growth Rates
【时间】2012年12月12日 (周三) 15：30-17：00
【地点】上海财经大学经济学院楼710室
【语言】英文
【摘要】This paper investigates the effect of high growth rates seen in the data on the Hilbert spectrum. This paper proposes three normalization scheme as an alternative of gain transformation: the EMD based normalization, the running-mean based normalization and the nonlinear iterative EMD based normalization approaches. Spectral analysis results of annual U.S. GDP series from 1790 to 2009 indicate that the straightforward EMD implementation on non normalized data could extract the short term variations, but lacks the information of the longer term changes due to the missing hidden scale components. Among the three approaches, the one that performs the best is the nonlinear iterative EMD approach. The advantage of this new approach is precisely to provide the data to study the longer term cycles that are impossible from the gain. Thus this new approach opens up a new alternative for the cycle studies directly.
