均值和方差双重变点的贝叶斯侦测

Bayesian Detection of Structure Changes of both Mean and Variance

本文应用贝叶斯方法研究了股价时序的均值和方差双重变点问题。基于后验概率比,我们提出一个类似ICSS算法的快速侦测算法。通过对上证指数时序的实证分析,我们总共发现5处方差突变。其中,3处是均值和方差双重变点,它们都对应中国股市的重大结构变化。

This article uses a Bayesian procedure to study structure changes of both mean and variance in stock price time series.A fast algorithm like ICSS algorithm is proposed to detect change points of both mean and variance based on posterior odds.We discover five changes of variance in a Shanghai composite index time series.Three of them are change points of both mean and variance that indicate significant structure changes in Chinese stock market.