方差无穷厚尾序列持久性变点的截尾检验  

Truncated Test for Persistence Change Points under Heavy-tailed Sequences of Infinite Variance

作　　者：夏小刚[1] 鲁珍 Xia Xiaogang;Lu Zhen(Xi’an University of Science and Technology)

机构地区：[1]西安科技大学

出　　处：《哈尔滨师范大学自然科学学报》2020年第2期11-18,共8页Natural Science Journal of Harbin Normal University

基　　金：国家自然科学基金(71473194,71273206);陕西省教育厅科学研究计划资助项目(16JK1500);陕西省科技厅自然科学基金(2013KJXX40,2017JM1024)。

摘　　要：基于分母为残差平方累积和的Ratio统计量文章对方差无穷厚尾序列的持久性变点进行截尾检验.截尾后,原假设下统计量的渐近分布收敛于维纳过程,与厚尾指数基本无关,备择假设下统计量具有一致性.Monte Carlo模拟结果表明截尾检验对经验水平没有显著的影响,但对经验势有明显的提高,说明了截尾方法的有效性.A truncated test based on a ratio statistic that the denominator is the cumulative sum of residual squares to study the persistence change point problem of heavy-tailed observations with infinite variance in this paper.After the truncation,the asymptotic distribution of the statistic under the null hypothesis is converged to the Wiener process,which is basically independent of the thick tail index.The statistic under the alternative hypothesis is consistent.Monte Carlo simulation results show that the truncated test has no significant effect on the empirical size,but the empirical power is obviously improved,indicating the effectiveness of the truncated method.

关 键 词：截尾检验 持久性变点 Ratio统计量 厚尾序列 

分 类 号：O212.1[理学—概率论与数理统计]