多元分布下单边备择假设的两步检验  

A Two-step Test with the"One-side"Alternative in Multinomial Distribution

作　　者：谢民育[1] 吴茗[1] 熊明[1] 

机构地区：[1]华中师范大学数学与统计学学院,湖北武汉430079

出　　处：《数理统计与管理》2011年第2期227-233,共7页Journal of Applied Statistics and Management

基　　金：国家自然科学基金资助项目(项目编号10571070)

摘　　要：x^2检验是熟知的拟合优度检验.它适用于任意的备择假设,尤其是适用于一般的备择假设"H_0不真"的情况.但是,当备择假设是一个单边假设时,x^2检验没有用到单边假设所提供的信息.因此,一个充分利用单边备择假设信息的检验是有可能优于传统的x^2检验的.本文就是利用单边备择假设的信息给出了一个两步检验,并通过一个例子,详细介绍了两步检验法的实施,说明了其应用.The chi-square test is a well-known goodness-of-fit test.It is available for arbitrary alternative hypothesis,particularly for a general alternative＂H_0 is not true＂.However,when the alternative is a＂one-side 8＂hypothesis,the chi-square test does not use the information offered by the one-side hypothesis. Therefore it is possible that an appropriate test,which uses the information of the one-side alternative, may be better.This note gives a two-step test with the information offered by the one-side alternative. We describe in detail the procedure of two-step test method,and show with an example how it is used in a one-side alternative hypothesis test.

关 键 词：x2检验 单边备择假设 两步检验 

分 类 号：O212.4[理学—概率论与数理统计]