时间序列的图模型及其在股市相关性中的应用

The Graphical Model for Multivariate Time Series and Its Application to the Correlation of Stock Markets

图模型方法是高维数据统计分析的重要工具,时间序列的图模型方法有链图、因果图和偏相关图,将基于VAR模型的时间序列链图和因果图应用于国际股票市场,研究主要股指的动态相关性,结果表明:美国股市对周边股市的影响较大。将偏相关图应用于亚洲股票市场,研究亚洲主要股指的交互作用,结果表明:中国内地是相对独立的市场,中国香港、台湾以及新加坡、日本股票市场之间存在显著的信息流动。

We introduce the graphical models for multivariate time series： chain graph,causality graphs and partial correlation graphs.The chain graph and causality graphs based on vector autoregressive regression（VAR） model are applied to the international markets to examines the dynamic interrelaionships,results show that SP500 index of American strongly affect other index.The partial correlation graphs is applied to the Asian stock markets to find the interactions structure between the stock markets,Empirical results show that Chinese＇s stock market is independent,while that of Chinese＇s Hongkong and Taiwan,Japan,Singapore is strongly connected.