基于三种相关系数的时变Copula模型的比较及应用  

Comparison and Application of Time Varying Copula Models Based on Three Correlation Coefficients

作　　者：郎诩森 何坤[1] LANG Xusen;HE Kun(College of Science, Donghua University, Shanghai 201620, China)

机构地区：[1]东华大学理学院,上海201620

出　　处：《东华大学学报（自然科学版）》2018年第6期1014-1018,共5页Journal of Donghua University(Natural Science)

基　　金：国家自然科学基金资助项目(NSFC11301068)

摘　　要：针对股票市场的波动性及非正态分布的特点,以t-GARCH(generalized autonegressive condition heteroskedasticity)模型描述其收益率的波动性并建立边缘分布,再分别从线性相关系数、Kendall秩相关系数、Spearman相关系数出发建立时变正态Copula模型。基于该模型描述变量间的相关结构,并通过信息准则比较几种模型的优劣。股票数据的实证检验与分析表明,基于Kendall秩相关系数建立的时变模型更能捕捉股票市场的相依机制。According to the characteristics of volatility and non-normal distribution of the stock market,this paper describes the volatility of stock-market yield with the t-GARCH(generalized autonegressive condition heteroskedasticity)model and establishes the marginal distribution.Then the time-varying normal Copula model is established based on the linear correlation coefficient,Kendall rank correlation coefficient and Spearman correlation coefficient.The correlation structure is described between variables with the Copula model and the advantages and disadvantages of these models are compared using information criterion.On the basis of the empirical test and analysis of the stock data,it shows that the time-varying model based on the Kendall rank correlation coefficient is most suitable for the stock market.

关 键 词：股票市场 波动性 t-GARCH 相关系数 时变Copula模型 

分 类 号：O211.1[理学—概率论与数理统计] F830[理学—数学]