极值理论和贝叶斯估计在VaR计算中的应用  

The application of Extreme value theory and Bayesian estimation for the VaR calculation

作　　者：熊健[1] 林煌[1] 

机构地区：[1]广州大学数学与信息科学学院,广东广州510006

出　　处：《广州大学学报（自然科学版）》2013年第3期5-8,共4页Journal of Guangzhou University:Natural Science Edition

摘　　要：利用广义GPD分布拟合香港恒生指数日收益率的尾部分布,采用贝叶斯方法对模型的参数进行估计,这样既能充分利用先验信息又能有效地融合样本数据;之后与极大似然法相比较,结果表明,使用贝叶斯方法在较高的置信度得到的VaR大于使用极大似然方法所得,但精确度还是有待提高.这说明极大似然方法将低估风险,提醒人们需要寻找恰当方法,进一步提高风险意识.In this paper,we apply the General Pareto Distribution（GPD） on fitting the tail of distribution of the return of Hang Seng Indices. A Bayesian method is employed to estimate the parameters of model. It not only makes good use of the prior information but also combines with the sample data effectively. Compared with MLE method, the Bayesian method has a greater VaR value in a high degree of confidence. This shows that the MLE method will be underestimated risk and the risk managers need to look for an appropriate approach for improving their risk awareness.

关 键 词：贝叶斯 极值理论 在险价值 

分 类 号：O212[理学—概率论与数理统计]