Multiple testing
Benjamini–Hochberg (FDR)
Implemented
multiple-testing.benjamini_hochbergDefinition
Controls the expected proportion of false discoveries among rejected hypotheses.
Formula
Reject H₍₁₎…H₍ₖ₎ for the largest k with p₍ₖ₎ ≤ (k / m) α
Inputs and outputs
- p_values
- alpha
Returns: adjusted p-values, reject flags
Assumptions
- Independent or positively dependent test statistics.
Limitations
No metric-specific limitations are documented yet. Interpret the value alongside the task, data, and other metrics.
Python API
import evalsuite as es
es.adjust_pvalues(p_values, method="bh")References
- Benjamini, Y., & Hochberg, Y. (1995). Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the Royal Statistical Society: Series B, 57(1), 289–300.