附加噪声长记忆过程的半参数估计方法研究

附加噪声长记忆过程的参数估计在实证研究中一直受到回避.本文通过聚合算子对样本数据进行降噪处理,研究了局部Whittle(LW)估计和对数周期图(LP)回归两种半参数估计方法.结果表明LW估计方法相比LP回归,解决了半参数方法的参数选取问题,且能够完全忽略噪声数据的影响而得到一致的估计.将LW估计应用于中国股市,发现重大突发事件发生时的长记忆性表现得最为强烈,且事件后比事件前要强烈.

Semi-Parametric Estimation Methods on Long-Memory Processes with Added Noise

The parameter estimation of long memory process with added noise is always ignored.In this work,after the sample data are aggregated,the methods of semi-parametric estimation including local Whittle estimation and log-periodogram regression are used.Compared to LP regression,LW estimation was proved to solve the parameter selection,and it also can neglect the impression of noise data.Then LW estimation was applied in China stock market.The results presented that long memory was found greater during the gravely outburst events,and which was stronger in the post-event than in the pre-event.