双边数据风险差的一致性检验

Homogeneity Test of Risk Difference for Bilateral Data

本文探讨了多组双边数据风险差的一致性假设检验问题及其检验过程,在一致性检验中,当样本量较小时,Wald统计量和Score统计量的第一类错误率接近于预设的显著性水平0.05,然而似然比统计量显示出了比较膨胀的第一类错误率。当样本量较大时,Score统计量检验效果更佳。随着样本量的增加,所有统计量检验效果趋于稳健。因此,在评估第一类错误率性能时,对于多组双边数据,建议采用统计量Score统计量进行风险差的一致性检验。

This article explores the homogeneity testing of risk difference for multiple sets of bilateral data and its testing process. In the homogeneity test, when the sample size is small, the Type I error rates of Wald statistic and Score statistic are close to the preset significance level of 0.05, while Likelihood ratio statistic exhibits a relatively inflated Type I error rate. However, when the sample size is larger, the testing effect of Score statistic is better. As the sample size increases, the testing effects of all statistical measures tend to become more robust. Therefore, when evaluating the performance of Type I error rates for multiple sets of bilateral data, it is recommended to adopt Score statistic for the homogeneity test of risk difference.