{"entity":{"id":"false-discovery-rate","kind":"term","name":"False discovery rate","aka":["FDR","Benjamini-Hochberg procedure","BH procedure","q-value","FDR-adjusted p-value"],"tldr":"The false discovery rate is the expected fraction of false findings among the findings a procedure selects, counted as zero when nothing is selected; it describes the shortlist as a set, not any one result.","summary":"False discovery rate is the expected proportion of false rejections among all rejections made by a multiple-testing procedure, taking the proportion as zero when none are made. It describes repeated use of that procedure under its assumptions. It is not the probability that each selected finding is false, a guaranteed false fraction in one list or evidence of biological importance or patient benefit.","asOf":"2026-11-11","links":[{"label":"Benjamini and Hochberg: controlling the false discovery rate (Journal of the Royal Statistical Society B 1995)","url":"https://doi.org/10.1111/j.2517-6161.1995.tb02031.x"},{"label":"Benjamini and Yekutieli: false discovery rate control under dependency (Annals of Statistics 2001)","url":"https://doi.org/10.1214/aos/1013699998"}],"tags":[],"related":["p-value","statistical-significance","ppv","preclinical","prespecified-vs-post-hoc"],"cancers":[],"sections":[],"technologies":[],"targets":[],"drugs":[],"companies":[],"institutions":[],"pathways":[],"terms":[],"trials":[],"people":[],"bottlenecks":[],"keyPapers":[],"journals":[],"dependsOn":[],"notes":["Write V for false rejections and R for all rejections: the false discovery proportion is V/R, defined as zero when R is zero, and FDR is its expectation. Family-wise error is the probability of at least one false rejection; the two coincide when every null hypothesis is true but generally differ otherwise. Benjamini-Hochberg sorts all m valid p-values, compares rank i with (i/m)q, finds the largest passing rank and selects everything through it. The original control result assumes independence; positive-dependence extensions and procedures for broader dependence have their own conditions. Correlation is not permission to ignore the procedure's assumptions.","Fictional Benjamini-Hochberg examples use target q = 0.05. For five sorted p-values 0.001, 0.008, 0.026, 0.080 and 0.200, rank bounds are 0.01, 0.02, 0.03, 0.04 and 0.05; rank 3 is the largest passing rank, selecting the first three. For three values 0.018, 0.030 and 0.060, bounds are about 0.0167, 0.0333 and 0.05; rank 1 fails its individual comparison but rank 2 passes, so the first two are selected. Do not stop at the first failure. The two illustrative plots use different labelled vertical ranges and test counts. Selection assigns neither finding a five per cent false probability.","Keep the complete hypothesis family and valid p-values, target, procedure and dependence assumptions, then report the selected set or adjusted values under that method. Correcting separate families and combining their hits does not automatically control the combined list. Q-values and adjusted p-values can name different calculations, so identify the actual method. Gene-, peptide- and protein-level identification claims also have different units; control at one level does not automatically transfer to another. The fraction of diagnostic positive calls that are false, one minus PPV, is a different quantity from the expected false-rejection fraction of a multiple-testing procedure. Omics selection prioritizes further investigation; it does not identify which hit is false or establish clinical utility."],"category":"Methods and models"},"route":"/terms/false-discovery-rate/","neighbours":{"term":[{"id":"p-value","kind":"term","name":"P-value","route":"/terms/p-value/"},{"id":"ppv","kind":"term","name":"Positive predictive value (PPV)","route":"/terms/ppv/"},{"id":"prespecified-vs-post-hoc","kind":"term","name":"Pre-specified vs post-hoc analysis","route":"/terms/prespecified-vs-post-hoc/"},{"id":"preclinical","kind":"term","name":"Preclinical","route":"/terms/preclinical/"},{"id":"statistical-significance","kind":"term","name":"Statistical significance (P values, alpha, multiplicity)","route":"/terms/statistical-significance/"}]}}