基于2个独立二项分布的风险比的信仰推断  

Fiducial Hypothesis Testing for the Ratio of the Parameters of Two Independent Binomials

作　　者：汪永辉[1] 金华[1] 吴琴[1] 

机构地区：[1]华南师范大学数学科学学院,广州510631

出　　处：《华南师范大学学报（自然科学版）》2016年第1期114-118,共5页Journal of South China Normal University(Natural Science Edition)

基　　金：国家自然科学基金青年项目(11401226)

摘　　要：2个独立二项分布参数之间的风险比的非劣效性检验在医学统计研究中是一个非常有意义的问题,常用的限制性极大估计方法在大多数情况下都不能控制第一类错误.文中提出用基于信仰推断法来解决基于2个独立二项分布参数之间的风险比的非劣效性检验问题.模拟结果显示:在小样本的研究情况下,这种基于信仰推断法的MF检验方法能很好地控制第一类错误,检验功效也不差.In medical statistics study, non-inferiority test for two independent binomial distribution parameters is a very important problem. The constrained maximum likelihood test statistic cannot control the type I error rates for some cases be investigated. In this article, the fiducial inference methodology is used in order to develop more pow- erful tests for non-inferiority based on the ratio between two independent binomial distributions. A broad Monte Carlo comparison between different tests for non-inferiority is presented, confirming the preference of the proposed method from a power perspective. Simulation studies suggest that the MF test can control the type I error rates and its empirical type I error rate are much closer to the prespecified nominal significance level than those of other tests well with competitive powers.

关 键 词：二项分布 非劣性检验 信仰推断 P值 功效 

分 类 号：O213.2[理学—概率论与数理统计] O213.8[理学—数学]