期刊名称：Brazilian Journal of Probability and Statistics
印刷版ISSN：0103-0752
出版年度：2003
卷号：17
期号：1
页码：75-90
出版社：Brazilian Statistical Association
摘要：Statistical inference based on the normal model is known tobe vulnerable to outliers. Despite this fact and the considerable interest inrobust procedures in the statistical literature, most applied statistical analysiscontinues to be based on the normal model. Our approach is to replace thenormal model by a general lo cation-scale family of nonlinear mo dels whichinclude several asymmetric distributions that have a wide range of practicalapplications for analysing univariate data. We focus on the second-order cor-rections to the likelihood ratio and score statistics, since they are the mostcommonly used large sample tests. We obtain simple formulae for the correc-tions in some special location-scale models. We use Monte Carlo simulationto show that the corrected likelihoo d ratio and score tests have empirical sizescloser to the nominal sizes than the classical uncorrected tests even when thescale parameter in replaced by a consistent estimate
关键词：Asymptotic distribution; Bartlett correction; Bartlett-type;correction; likelihoo d ratio statistic; location-scale model; maximum likeli-;hood estimate; score statistic
