主 题：On the Seasonal Duration Adjustment of High-Frequency Data
主讲人：Wu Zhengxiao 博士
主持人：兰伟
时 间：2014年5月26日14:00-15:00
地 点：通博楼B212学术会议室
主办单位：统计学院 科研处
内容提要：
In the last decade, intensive studies on modeling financial data at the transaction level have been conducted. It has created a new body of literature called ``high-frequency finance''. In the analysis of high-frequency data, it is often the first step to adjust the data to remove the seasonal effect. Currently the most popular deseasonalization procedure is the cubic spline procedure proposed by Engle and Russell (1998). In this article, we first carry out a simulation study and show that the cubic spline procedure is not effective. Then we define seasonal point processes rigorously and prove a time change theorem. Based on which, a new seasonal adjustment procedure is proposed and its effectiveness is demonstrated. We argue that the clear theoretical interpretation, improved deseasonalization performance and the ease of implementation of the new approach recommend its use over the cubic spline procedure in virtually all applications.
