相关系数与相关性度量

研究了度量相关性的两个主要工具:线性相关系数和尾部相关系数.线性相关系数反映了变量间的线性相关性,这对于一般的椭圆型分布是合适的.但如果随机变量具有不对称的尾部变化特征时,要用尾部相关系数描述它们之间的相关性.通过相关函数C opu la,对沪深股市的尾部相关系数进行了定量分析.结果表明:沪深股市具有较强的相关性.

Research on Linear Correlation and Dependence Measure

Linear correlation and tail-dependence coefficient are emphasized in this paper.Linear correlation is a natural dependence measure for normal multivariate and,more generally,elliptic distributions.The concept of tail-dependence describes the amount of asymptotic dependence in the lower-left-qudarant tail or upper-right-qudarant tail of a bivariate distribution.An important measure for tail-dependence is given by the so-called tail-dependence coefficient.After presenting the statistical properties we provide an example and illustrate the tail dependences between Shanfhai and Shenzhen stock markets based on the optimal copula function.