False discovery rate control is not always a replacement for Bonferroni adjustment (Letter commenting on: J Clin Epidemiol. 2014;67:850-7)

What is it about?

A previous article suggested that when multiple comparisons are conducted in health research, false discovery rate control should generally be used rather than the Bonferroni procedure. However, that recommendation was based on misunderstandings. The present letter identifies some of those misunderstandings. False discovery rate control and the Bonferroni procedure perform different "jobs," so neither approach is universally better than the other for all situations.

Featured Image

Why is it important?

Misconceptions about Type I error control are common, and it is important to correct those misconceptions when they appear in published articles. For example, the myth that the Bonferroni procedure only addresses the "universal null hypothesis" has long been perpetuated in the epidemiology literature (and elsewhere). It is also important to note that although false discovery rate control is useful for some situations, it has important limitations that make it inadvisable for other situations.