{"id":"arxiv/2608.02507","url":"https://arxiv.org/abs/2608.02507","tags":["math.ST","cs.IT","cs.LG"],"score":0,"title":"Beyond Modern Asymptotics for Log-Likelihood Ratios in Logistic Regression","domain":"math.ST","source":"arxiv","authors":["Hugo Chardon","Reese Pathak","Nikita Zhivotovskiy"],"reviews":0,"subject":"math.ST","summary":"We characterize the finite sample behavior of the log-likelihood ratio statistic in binary logistic regression, uniformly over both the design and the target parameter. For $n\\geq d\\geq 3$, we determine, up to universal constants, its worst case $(1-δ)$ quantile over all fixed collections of design vectors and all target parameters: \\[ d\\log\\left(\\frac{e n}{d}\\right)+\\log\\left(\\frac{1}δ\\right). \\] This is a nonasymptotic analogue of the Wilks $χ^2_d$ phenomenon and requires no regularity assumptions on the design. The low dimensional cases exhibit unusual behavior. The worst case quantile in…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-08-03","reactions":{}}