False Discovery Rate

The Benjamini–Hochberg procedure creates an ordered comparison rather than evaluating each p-value in isolation. Researchers rank the p-values from smallest to largest, then evaluate them against thresholds that depend on their rank within the tested set and the selected significance level. This links the decision rule to the number and ordering of tested hypotheses.

Rank position matters because the threshold assigned to a p-value reflects where it falls among all tested hypotheses. Consequently, the decision is shaped by the full collection of results, not only by one outcome. This feature allows a multiple-comparison analysis to retain sensitivity while still applying a formal criterion to the discovery set.

An FDR result should not be read as proof that every selected finding is genuine. Its criterion concerns the expected proportion of false positives among the findings classified as significant. In psychology, that distinction is important when interpreting a group of outcomes from one study, because control operates across the discovery set rather than guaranteeing each result.

FDR differs from a strategy designed to eliminate nearly every potential false positive. By accepting control of an expected false-discovery proportion, it can preserve greater sensitivity when many hypotheses are tested. This trade-off is especially relevant when a study examines numerous behavioral or cognitive outcomes and would otherwise risk overlooking potentially informative findings.

To apply the Benjamini–Hochberg approach, assemble the p-values from the hypotheses included in the multiple-comparison analysis, rank them, and compare each with the threshold associated with its position and chosen significance level. The resulting comparisons determine which hypotheses are treated as significant under the procedure, providing a consistent basis for interpreting the tested set.

False Discovery Rate is particularly useful when an investigation produces many simultaneous tests rather than a single primary comparison. Examples in psychology include cognitive experiments, surveys, neuroimaging analyses, and large-scale behavioral research. In these settings, applying one multiple-comparison criterion helps researchers evaluate a broad collection of outcomes while maintaining sensitivity to discoveries.

In neuroimaging and behavioral studies, the number of analyzed outcomes can make significance decisions difficult to interpret without a multiple-comparison framework. FDR supplies a way to evaluate those results as a set, which is why it is relevant to psychology research spanning brain-related measures, cognitive performance, survey responses, and other behavioral observations.