基于MOSUM的合并带宽变点估计方法

MOSUM-based Merged Bandwidth Variable Point Estimation Method

在变点的应用研究中,针对时间序列中存在短时间间隔内大变化和长时间间隔内小变化的变点问题,现有的检测方法都难以解决。文章以此为出发点做进一步的探究,首先,基于移动和统计量(MOSUM),考虑了一个可能存在时间序列误差的多均值变点模型,并构造检验统计量。其次,使用移动和类型估计器对长期运行方差估计,并考虑带宽对检验统计量的影响,提出了合并带宽MOSUM检测方法(MMT)。最后,通过对该方法进行模拟比较,实证分析了美国Nasdaq股市在2005年1月至2015年3月的收益率变化情况。结果表明:在检测准确率和精度上,MMT方法整体比WBS、cumSeg和PELT方法更好,同时也能准确识别Nasdaq股市收益率产生突变的时间点。

In the application research of variable point,the existing detection methods are difficult to solve the problem of variable point with large change in short time interval and small change in long time interval in time series.This paper takes this as a starting point to make a further exploration.First,based on the moving sum statistics(MOSUM),the paper considers a multimean variable-point model with possible time series errors to construct a test statistics.Secondly,using moving sum type estimator to estimate the long run variance,and considering the effect of bandwidth on the test statistics,the paper proposes a merged bandwidth MOSUM test method(MMT).Finally,through the simulation comparison of the method,the paper empirically analyzes the yield change of the Nasdaq stock market in the United States from January 2005 to March 2015.The results show that the MMT method is better than WBS and cumSeg and PELT method in terms of detection accuracy and precision,and it can also accurately identify the time point at which the Nasdaq stock market yields have taken a sudden turn.