高维数据下关于正态均值线性约束的检验  

Test of linear constraint of normal mean value in high dimensional data

作　　者：李熠璇 范永辉[1] LI Yi-xuan;FAN Yong-hui(School of Mathematical Science,Tianjin Normal University,Tianjin 300387,China)

机构地区：[1]天津师范大学数学科学学院,天津300387

出　　处：《哈尔滨商业大学学报（自然科学版）》2021年第1期73-78,共6页Journal of Harbin University of Commerce：Natural Sciences Edition

基　　金：天津市教育委员会科研项目(JW1714).

摘　　要：高维数据经常会出现数据维数远远大于样本容量的现象,导致多元统计分析中许多经典检验方法失效,为弥补这一缺陷,以似然比检验方法为基础,结合并交原则,对于正态总体均值受线性约束这种假设检验问题提出了新的广义似然比检验方法,并通过数据模拟方法验证该检验与Bai和Saranadasa的检验方法相比能更好的控制检验水平.High dimensional data often have the phenomenon that the data dimension is much larger than the sample size,which leads to the failure of many classical test methods in multivariate statistical analysis.In order to make up for this defect,based on the likelihood ratio test method,combined with the union-interection principle,a new generalized likelihood ratio test method was proposed for the hypothesis test problem that the normal population mean was linearly constrained,which was verified by the data simulation method compared with Bai and Saranadasa,the new test can control the test level better.

关 键 词：高维数据 并交原则 广义似然比检验 线性约束 显著水平 蒙特卡洛方法 

分 类 号：O212.4[理学—概率论与数理统计]