{"id":"arxiv/2610.38763","url":"https://arxiv.org/abs/2610.38763","tags":["math.ST"],"score":0,"title":"Central limit theorems for high-dimensional ridge regression under uniform subsampling","domain":"math.ST","source":"arxiv","authors":["Thomas Lam","Nina Dörnemann","Holger Dette"],"reviews":0,"subject":"math.ST","summary":"We study ridge regression based on uniformly subsampled observations in high-dimensional linear models, where the number of predictors and the sample size grow proportionally. Subsampling is implemented through independent Bernoulli indicators with a common selection probability. We establish central limit theorems for linear contrasts of the ridge estimator, allowing for non-Gaussian predictors and general population covariance matrices. Under a vanishing-signal assumption, we obtain an explicit bias correction in the isotropic case and develop consistent estimators of the asymptotic…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-28","reactions":{}}