基于Bootstrap方法的多组配对数据风险差的一致性检验

Homogeneity test based on Bootstrap method of risk difference for multi-group bilateral data

在Dallal模型下构建3个统计量,研究小样本多组配对数据风险差的一致性检验问题.通过Monte-Carlo数值模拟,比较小样本下渐近方法和Bootstrap方法在检验过程中的性能.研究发现,Bootstrap方法在第一类错误率和功效上表现更优;特别地,基于Score检验的Bootstrap方法的第一类错误率较稳健,且有较高的功效,检验性能最优.

Under the Dallal model,three statistics are constructed to study the homogeneity test of risk differences for small sample multi-group bilateral data.Compared the performance in statistical test process of asymptotic method and Bootstrap method with the small sample case by means of Monte-Carlo simulation,the result shows that the Bootstrap method performs better in the empirical type I error rate and power.In particular,the Bootstrap method based on the Score test exhibits a relatively robust type I error rate,high power,and optimal testing performance.