完全随机设计两样本率比较的非条件确切检验方法  

Unconditional Exact Test for Comparing Two Binomials in Independent Samples

作　　者：韩宏[1] 王彤[1] 何大卫[1] 

机构地区：[1]山西医科大学卫生统计教研室,030001

出　　处：《中国卫生统计》2005年第4期207-209,共3页Chinese Journal of Health Statistics

摘　　要：目的探讨完全随机设计两样本率比较的Barnard确切非条件检验方法,并与Fisher确切概率法进行比较。方法非条件确切检验方法是通过使冗余参数在其参数空间内变动,以其确切P值的上确界为所求检验P值。结果根据实例分别算得两种方法的确切分布与P值,Barnard方法所得P值较Fisher方法更小。结论在小样本情形下,Barnard确切非条件检验法较Fisher确切条件检验法获得的检验P值更小,检验功效更高。Objective To explore the Barnard＇s unconditional exact test for comparing two binomials in independent samples, and to compare it with Fisher＇s method. Methods In unconditional exact test methods, all possible values of nuisance parameter are searched through its parameter space, and define the maximum exact P - value as final P - valuz.Results Real example was calculated with two methods, and P-values of Barnard＇s method were smaller than Fisher＇s method. Conclusion In inference with small samples, one can get smaller P-values and higher power with Barnard＇s method than Fisher＇s method.

关 键 词：四格表 确切非条件检验 Barnard检验 基本原理 卫生统计 

分 类 号：O211.6[理学—概率论与数理统计]