针对股市收益分布的“尖峰肥尾”特征,引入了偏t分布作为新息分布。基于VaR方法,从风险估计的角度,利用ARFIMA(2,d_1,0)-HYGARCH(1,d_2,1)-skt模型对1996年12月17日至2007年7月5日期间的沪深股市收益进行了实证分析.实证结果显示:沪深股市具有显著的双长记忆特征;上海股市的日收益率和波动率的长记忆性均比深圳股市强;ARFIMA(2,d_1,0)- HYGARCH(1,d_2,1)-skt模型对我国股市收益具有较强的风险估计和预测能力。 In view of the “peak fat tail ” characteristic of the stock market return distribution, the partial t distribution is introduced as the new interest distribution. Based on the VaR method, from the perspective of risk estimation, using the ARFIMA (2, d_1,0) -HYGARCH (1, d_2,1) -skt model, the stock market of Shanghai and Shenzhen Stock Exchanges from December 17, 1996 to July 5, 2007 The empirical results show that: the Shanghai and Shenzhen stock markets have significant dual long memory characteristics; Shanghai stock market daily returns and volatility of the long memory are stronger than the Shenzhen stock market; ARFIMA (2, d_1,0) - HYGARCH (1, d_2,1) -skt model has strong risk estimation and prediction ability to the stock market returns in China.