持久性变点的非参数检验  

A Nonparametric Test for Change in Persistence

作　　者：秦瑞兵[1] 田铮[1,2] 金浩[1] 

机构地区：[1]西北工业大学应用数学系,陕西西安710072 [2]中国科学院遥感应用研究所国家遥感科学重点实验室,北京100101

出　　处：《西北工业大学学报》2010年第2期304-307,共4页Journal of Northwestern Polytechnical University

基　　金：国家自然科学基金(60972150;10926197;60375003);西北工业大学科技创新项目(2007KJ01033)资助

摘　　要：文章基于符号函数给出一种检验持久性变点的统计量。同时,在原假设下得到了其渐近分布,在备择假设下证明了该检验的容许性。数值模拟表明,文中所给检验在新息厚尾时具有较好的势,且犯第1类错误的概率较小。最后将文中结果应用到美国联邦季度预算赤字数据中,表明所给方法的有效性。This paper considers testing against a change characterized by a shift in persistence of a time series,namely,a time series shift from stochastic stationarity to stochastic nonstationarity.This problem has been well documented by Refs.2 through 7,but when the innovations are heavy-tailed series,the tests proposed in Refs.2 through 7 have poor performances,so in this full paper,a nonparametric test is proposed to improve the power and the size of the tests.Section 1 deals with the proposed variance ratio test based on the indicators of the data minus the sample median and derives its asymptotic distribution under the null hypothesis and its consistency under the alternative hypothesis.Monte Carlo simulations in Section 2 demonstrate that the proposed test has better power and size than those tests in Refs.2 through 7 when the innovations are heavy-tailed.An application to the U.S.government budget deficit shows the validity of the proposed test in Section 3.

关 键 词：平稳过程 单位根过程 持久性 变点 协整系统 

分 类 号：O212[理学—概率论与数理统计]