变系数模型的变量选择及在股票数据中的应用

Variable selection of varying-coeffcient models and its application on stock data

作者研究了纵向数据分析中变系数模型的变量选择及效应估计问题,该模型允许变量的效应随时间改变.本文方法在进行变量选择的同时,也估计变系数函数,避免了传统的变量选择方法极其复杂的计算.将本文方法用于股票价格分析,能够快速地在公司的众多财务变量中挑选出对股票收益率有显著影响的变量,并估计这些变量的时变效应,很好地解释股票收益率的变化.

This paper discusses the variable selection and estimation based on varying-coeffcient models for longitudinal data.The model allows the effect of variables to vary with time.The method in this paper estimates the functions of varying-coeffcient and selects variables simultaneously,which avoids the intensive computation for the traditional variable selection.Applying this method to stock price,the variables are selected quickly which have significant effect on the return rate of stock from the numerous company financial variables,and the time-varying effect of those significant variables could be estimated simultaneously.The results show that this method works well.