基于信息量准则的广义Lambda分布变点分析  

Change point analysis with generalized Lambda distribution based on information criterion

作　　者：黄传明[1] 张晓旭[2] 张三国[1] 

机构地区：[1]中国科学院大学数学科学学院,北京100049 [2]中国科学院大学管理学院,北京100190

出　　处：《中国科学院大学学报（中英文）》2013年第6期728-736,743,共10页Journal of University of Chinese Academy of Sciences

基　　金：教育部回国留学人员科研启动基金;中国科学院研究生院院长基金(Y05101AY00)资助

摘　　要：研究广义Lambda分布的变点问题,结合信息量准则方法实现了变点检测.前人曾指出,为使变点检测结果更具统计说服力,可引入显著性水平α及相关临界值cα.本文指出,在一般条件下,此时与通常极大似然比检验渐近结果是等价的.给出5种不同信息量准则下模拟数据的检测结果及分析,并将该方法应用于实际数据.实验结果表明,该方法能够有效识别出变点位置.In this article, we study the change point problem with generalized Lambda distribution based on information criterion. It was previously pointed out that, to make the conclusion more statistically convincing, the significant level α and its associated critical value cα could be introduced. We prove that, under ordinary conditions, this kind of information criterion is equivalent to the maximum likelihood ratio test in asymptotic formation. The simulation study is carried out with five different information criterions and the method is applied to treat practical data. The results show the validity of this method.

关 键 词：广义Lambda分布 变点检测 信息量准则 似然比检验 

分 类 号：O212.1[理学—概率论与数理统计]