多元响应回归模型的变点检测及其应用

Change Point Detection and Application of Multivariate Response Regression Model

文章基于多元T分布和施瓦茨信息准则(SIC)研究了多元回归模型的参数估计方法和变点检测问题。首先将多元T分布看成多元正态分布的混合分布,进而应用EM算法得到了模型的参数估计方法;然后,利用SIC信息准则提出了模型中变点识别方法。数值模拟分析进一步验证了所提的估计方法及其SIC准则变点检测方法的有效性。最后,应用SIC准则方法研究了上证指数数据的变点检测问题。

This paper is based on multivariate T distribution and Schwarz information criterion(SIC)to study the parameter estimation method and change point detection for multivariate regression model.Firstly,the paper takes the multivariate T distribution as the mixture of multivariate normal distribution and uses the EM algorithm to obtain the parameter estimation method of the model,and then employs SIC to propose the detection method for the change point in the model.Numerical simulation analysis further verifies the validity of the proposed estimation method and its SIC change point detection method.Finally,the paper uses SIC method to study the change point detection for the data of Shanghai securities composite index.