【摘要】 采用图形建模工具中的有向非循环图(DAG)方法对高维随机变量进行Pair-Copula分解,并提出非正态的PC算法对DAG进行识别,最后应用于国际股市分析主要股指的尾部相依结构。数值模拟表明,若变量不服从椭圆分布,新的非正态PC算法要优于传统的PC算法。经AR(1)-GARCH(1,1)模型过滤后的残差DAG分析,股市信息流动具有明显的区域效应,欧洲股市主要通过美国与美洲其它股指存在信息传导。英国和香港是欧洲和亚洲股市信息联系的纽带,中国内地只与香港存在直接的信息流动。基于SJC-Copula实证结果表明,欧洲德国和英国股市间的尾部相关性较强,亚洲股市间下尾相关系数要大于上尾,不同区域联结股市间的上尾相关性强于下尾。更多还原

【Abstract】 The Pair-Copula constructions based on directed acyclic graph which is distinguished by Nonnormal PC algorithm is proposed and used in the dependence structure among the international stock markets.Simulation shows that the Nonnormal PC algorithm is superior to the traditional method.Empirical result shows that the regional segmentation of the major international financial markets is proved in this study,the connection between America and European stock markets by US;the connection between European and Asian stock markets by England and Hong Kong;Chinese stock market is only connected Hong Kong directly.The tail dependence between German and England stock markets is the strongest the lower tail dependence between Asian stock markets is stronger than the upper,while the upper tail dependence between different continents is stronger than the lower.更多还原