{"id":"arxiv/2607.12208","url":"https://arxiv.org/abs/2607.12208","tags":["math.ST","cs.AI","stat.ME"],"score":0,"title":"The Benjamini--Hochberg Procedure Can Fail to Control the FDR for Correlated Two-Sided Gaussian Tests","domain":"math.ST","source":"arxiv","authors":["Edgar Dobriban"],"reviews":0,"subject":"math.ST","summary":"We show that the Benjamini--Hochberg procedure can fail to control the false discovery rate (FDR) at its nominal level for correlated two-sided Gaussian $p$-values. We construct a factor model for which, at level $α=0.01$, a rigorous interval-arithmetic certificate proves $FDR>0.0104$ for all sufficiently large numbers of hypotheses. This disproves a conjecture widely believed to be true for twenty years. Monte Carlo experiments are consistent with the theoretical result. The proof was obtained by GPT-5.6 Pro and carefully checked by the author.","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-07-13","reactions":{}}