基于Copula函数的秩相关和尾相关研究

Studies of Rank and Tail Dependence Represented by Copula Function

变量间的相关结构是考察变量联合分布特征的首要任务。最经典的是线性相关系数,然而由于线性相关系数是建立在正态分布假设基础上的,存在着很多的局限性,所以,引入了用Copula函数表示的秩相关和尾相关测度。用Copula函数表示的秩相关和尾相关测度比较稳健,不易受极端值影响,且不需要正态分布假设。最后通过模拟得到了上市公司t-Cop-ula的秩相关和尾相关测度。

To describe joint distribution＇ s characters, the first step is seeing about the dependent structure. The most classic one is the linear correlation coefficient. It is well known that the base of linear correlation coeglcient is normal distribution, so there are several limitations when using it. Rank and tail dependence represented by the Copula function are stable, and extreme values can＇ t affect them easily .They don＇ t need the assumption of normal distribution. A simulative example has been given to compute the t- Copula＇s rank and tail dependence.