沪深两市股票指数的长记忆性  

Long-term memory in Shanghai and Shenzhen stock markets

作　　者：姜仁娜[1] 叶俊[1] 

机构地区：[1]清华大学数学科学系,北京100084

出　　处：《清华大学学报（自然科学版）》2004年第12期1696-1699,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金资助项目(79970120)

摘　　要：针对中国股票市场的长记忆性问题,讨论了分整自回归移动平均(Auto-regressivefractionalintegratedmovingaverage,ARFIMA(p,d,q))模型中的参数估计问题,重点集中在对分整参数d的估计。使用Hurst指数方法估计d,并分别用经典R/S方法、有偏修正R/S方法和无偏修正R/S方法进行估计,并结合上证指数和深证成指的收益率数据,给出了3种方法的估计结果。实证结果表明,中国股票市场已初步显示出了长记忆性。给出ARFIMA模型的最优阶数和全部参数估计值。得出了上证指数和深证成指收益率所适合的最优的ARFIMA模型。The Auto-regressive Fractional Integrated Moving Average (ARFIMA(p,d,q)) model was used to analyze the return rates in the Shanghai and Shenzhen stock markets. The Hurst Exponent method was used to estimate the d parameter. The classical R/S method, the biased modified R/S method, and the unbiased modified R/S method were used to study the return rates of the Shanghai and Shenzhen stock markets. The results indicate that the Chinese stock markets show long-term memory effects. The paper gives the optimum ranks and estimates of all parameters in the ARFIMA(p,d,q) model for the optimum ARFIMA models for the return rates of the Shanghai and Shenzhen stock markets.

关 键 词：股票 概率论 ARFIMA(p d q)模型 长记忆性 R/S统计方法 

分 类 号：F830.91[经济管理—金融学] O211[理学—概率论与数理统计]