False Discovery Rate (FDR) and Benjamini-Hochberg
Hypothesis Testing
Balancing Discovery and Error Control
The Benjamini-Hochberg procedure controls the expected proportion of false discoveries, offering more power than Bonferroni for large-scale testing. It is the standard method for genomics, neuroimaging, and high-dimensional data.
- Genomics — Identifying differentially expressed genes while controlling false discoveries
- Neuroscience — Analyzing brain imaging data with millions of statistical tests
- Digital Advertising — Optimizing keyword testing at scale with FDR control
FDR control enables discovery without drowning in false positives.
The FDR is the expected proportion of rejected H₀s that are actually true (false discoveries):
The Benjamini-Hochberg (BH) procedure controls FDR at level q: