极值copula的统计推断及实证研究  

Statistical Inference for Extreme Value Copula with Empirical Research

作　　者：胡祥[1] 张连增[2] 

机构地区：[1]中南财经政法大学金融学院,湖北武汉430073 [2]南开大学金融学院,天津300071

出　　处：《数理统计与管理》2017年第1期151-161,共11页Journal of Applied Statistics and Management

基　　金：国家自然科学基金面上项目(71271121);国家社会科学基金面上项目(14BJY200)的资助

摘　　要：为了分析由极端事件所引起的巨额损失变量之间的相依关系,本文引入了比一般copula函数更有效的极值copula和上尾copula。我们介绍copula的对角截面以确定上尾相依系数。基于极大似然法,讨论了关于这些copula函数类的半参数估计方法。通过构建Cramer-Von Mises统计量对copula的拟合优度进行假设检验。在实证分析部分,我们通过具体的实例来说明,在应用研究中该如何选取最优的copula以描述变量之间的相关性。In order to anMyze the dependence between the catastrophe losses caused by extreme events, extreme value copulas and upper tail copulas are introduced, which play a more flexible role than copulas. We present the diagonal section of copulas for the purpose of determining the upper tail dependence coef- ficient. We study the semiparametric method of estimating these parametric copula families by maximum likelihood method. The goodness-of-fit testing procedures for them are also discussed by constructing Cramer-Von Mises statistic. In the empirical analysis, we illustrate the results by a numerical example, which tells us how to choose the fitted-best copula to model dependence in application studies.

关 键 词：极值copula 上尾copula 对角截面 半参数估计 假设检验 

分 类 号：O212[理学—概率论与数理统计]