Adaptive false discovery rate (FDR) procedures, which offer greater power than the original FDR procedure of Benjamini and Hochberg, are often applied to statistical maps of the brain. When a large proportion of the null hypotheses are false, as in the case of widespread effects such as cortical thinning throughout much of the brain, adaptive FDR methods can surprisingly reject more null hypotheses than not accounting for multiple testing at all-i.e., using uncorrected p-values. A straightforward mathematical argument is presented to explain why this can occur with the q-value method of Storey and colleagues, and a simulation study shows that it can also occur, to a lesser extent, with a two-stage FDR procedure due to Benjamini and colleagues. We demonstrate the phenomenon with reference to a published data set documenting cortical thinning in attention deficit/hyperactivity disorder. The paper concludes with recommendations for how to proceed when adaptive FDR results of this kind are encountered in practice.
Bibliographical noteFunding Information:
We thank Xavier Castellanos, Eva Petkova, Catherine Sugar, Martin Lindquist and Jason Lerch for very helpful discussions; Yin-Hsiu Chen, for bringing a number of key references to our attention; and two anonymous referees, for suggesting a number of improvements in the manuscript. Philip Reisss research was supported in part by National Science Foundation grant DMS-0907017 and National Institutes of Health (NIH) grant R01 EB009744- 01A. Armin Schwartzmans research was partially supported by NIH grants 1R21 EB012177-01A1 and 1P01 CA134294-01 . The cortical thickness study was funded by NIH grant R01 DA016979 .
- Adjusted p-values
- Attention deficit/hyperactivity disorder
- Cortical thickness
- False discovery rate
- Multiple testing
ASJC Scopus subject areas
- Cognitive Neuroscience