基于subsampling重尾序列持久性变点检验  

Subsampling Testing for Persistence Change in Heavy-Tailed Series

作　　者：秦瑞兵[1] 刘洋[1] 

机构地区：[1]山西大学数学科学学院,山西太原030006

出　　处：《山西大学学报（自然科学版）》2017年第2期209-215,共7页Journal of Shanxi University(Natural Science Edition)

基　　金：国家自然科学基金(71501115);中国博士后科学基金(2013T60266)

摘　　要：研究新息为方差无穷重尾序列的持久性变点检验问题,为得到较好的经验水平值,构造了DF型比率统计量,得到其渐近分布。为避免估计重尾指数κ,应用subsampling方法确定渐近分布的临界值并论证了该方法的合理性。最后,Monte Carlo模拟说明统计量及subsampling方法的有效性。This paper deals with the detection of persistence change in the heavy-tailed series with the infinite-variance innovations.In order to get better empirical sizes,the DF ratio statistic is extended and the asymptotic distribution is obtained.Subsampling method is applied to calculate the critical value of the statistic without estimating the tail indexκand the validity of subsampling method is established.Moreover,Monte Carlo simulations demonstrate the validity of the statistic and subsampling method.

关 键 词：SUBSAMPLING 持久性变点 重尾 DF型比率统计量 

分 类 号：O29[理学—应用数学]