{"id":"arxiv/2605.03781","url":"https://arxiv.org/abs/2605.03781","tags":["math.ST"],"score":0,"title":"Safe and Sharp Honest Inference for Nonparametric Estimation via Empirical Bernstein Calibration","domain":"math.ST","source":"arxiv","authors":["Zihao Yuan","Sven Klaassen","Holger Dette"],"reviews":0,"subject":"math.ST","summary":"Constructing honest confidence intervals often depends on reliable bias control and accurate calibration based on asymptotic normality, including standard-normal and folded-normal calibration. Substantial progress has been made in correcting or controlling smoothing bias, including robust bias correction and bias-aware inference. We first show that, even after smoothing bias has been well corrected or controlled, asymptotic-normality-based calibration may still be a binding source of finite-sample undercoverage. Thus, the resulting intervals may struggle to achieve the minimax shrinkage rate…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-05-05","reactions":{}}