基于2个独立二项分布的风险比的信仰推断

两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. 本文提出用基于信仰推断法来解决基于两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.

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

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, we use the fiducial inference methodology in order to develop more powerful tests for Non-inferiority based on the ratio between two independent binomial distributions. We present a broad Monte Carlo comparison between different tests for non-inferiority, confirming the preference of the proposed method from a power perspective. Simulation studies suggest that our 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.