临床研究中,我们会经常遇到多重检验的问题。当同时检验多个假设时,如何控制犯Ⅰ类错误的概率,保证结果的准确性,是研究者面对的首要问题。然而未对多重检验进行校正的文章比例仍较大。本文从简单的单个检验假设出发,回答为什么需要进行多重校正,何时进行多重校正,同时介绍总体错误率(FWER)和错误发现率(FDR)两种重要的错误率以及在此基础上的Bonferroni和Benjamini-Hochberg校正方法,从而避免因多重检验问题带来的混乱。

In clinical research, we often encounter the problem of multiple testing. When testing many hypotheses at the same time, how to control the type I error to ensure the accuracy of the results is the primary problem faced by researchers. However, the proportion of articles that didn't correct the multiple testing remains substantial. Starting with the simple hypothesis of a single test, this article provides an introduction to multiple testing issues, answers why and when multiple corrections are needed, introduces two important error rates which are family-wise error rate (FWER) and false discovery rate (FDR), and the Bonferroni and Benjamini-Hochberg correction methods based on them, thereby avoiding confusion caused by multiple testing.