【摘要】 L-矩方法现已成为金融领域厚尾分布分析和建模的重要工具,当极端值较多时,L-矩方法会变得较敏感,截L-矩不但具有L-矩方法的优良特征,而且增加了对首尾极端值的控制参数,对极端值存在的情况更加适用。论文采用AR(1)-GARCH(1,1)模型对收益率序列进行建模分析,得到近似独立同分布的残差序列。在此基础上,基于截L-矩考察广义帕累托分布对上证指数的损失尾部拟合情况,VaR的返回检验表明:基于截L-矩的GPD分布可以较好地拟合损失收益率的尾部,极值VaR可以有效地度量上海股市的风险。更多还原

【Abstract】 L-moment method has become an important tool in heavy-tailed distribution analysis and modeling.When extreme value is too much,L-moment method will become more sensitive.The trimmed L-moments(TL-moment) not only has excellent features analogously to L-moments,but also adds parameters to control extreme value sample of size,and it will be more applicable in this situation.In this papers,we use the AR(1)-GARCH(1,1) models to analyze return series and get residual series with approximate independent distribution.Trimmed L-moments are employed to identify the generalized Pareto distribution(GPD),and fitting GPD to a heavy-tailed data sample of the Shanghai index in the loss.Backtest of VaR shows that GPD based on the TL-moment can better fit behaviour of extreme tail,and extremum VaR can be an effective way to measure the risk of the Shanghai stock market.更多还原