摘要
应用极值理论计算风险值要求数据服从独立同分布的前提假设,但在实际的金融时间序列中并不满足这样的条件。为了解决该问题,可以引入极值指标,并利用除串法,重新对超阈值数据建立POT模型,对VaR和CVaR进行估计。利用美元兑人民币汇率数据进行实证分析,结果表明:利用除串法对超阈值的极值数据进行处理后,计算所得VaR和CVaR较未除串均有所降低,且在低置信水平(≤97.5%)下下降幅度比较明显;在高置信水平(≥99%)下运用失败率检验除串后的模型比较理想,并且较低的风险值能使用于缓冲损失的准备资金得到有效利用。
The application of EVT requires data to obey the independent and identically distributed assumption, but the actual financial time series are not satisfied with this hypothesis. To solve this problem, we introduce the extreme value index and build a POT model by using the method of declus- tering in order to calculate the estimate of VaR and CVaR. The computation result of USD/CNY fo- reign exchange rate proves that VaR and CVaR are lower after using the method of declustering, and the lowering is obvious at a low confidence level（〈=97.5%）, and the test of failure rate can be used at a high confidence level（〉=99%）. So the model after the use of declustering is an ideal one, and the lower VaR enables the preparation fund to be used effectively.
出处
《江苏大学学报（社会科学版）》
2015年第2期78-84,共7页
Journal of Jiangsu University(Social Science Edition)
基金
国家自然科学基金项目(11101364
11201421)
浙江省自然科学基金项目(Y6110110)
全国统计科研计划项目(2013LY137)
浙江省高校人文社科重点研究基地(统计学)资助项目
关键词
汇率
独立同分布
POT模型
除串
失败率检验
exchange rate
independent identically distribution
POT model
declustering
test of failure rate
