{"id":"arxiv/2606.25729","url":"https://arxiv.org/abs/2606.25729","tags":["stat.ME","math.ST","stat.CO"],"score":0,"title":"A Theory of Bootstrap Coverage Calibration for Generalized Posterior Credible Sets","domain":"stat.ME","source":"arxiv","authors":["Masahiro Tanaka"],"reviews":0,"subject":"stat.ME","summary":"Generalized posteriors replace the likelihood by an exponentiated empirical criterion, but their credible sets generally lack asymptotic justification for frequentist coverage. General posterior calibration selects a scalar learning rate by estimating coverage with the bootstrap. Using Edgeworth expansions under regular fixed-dimensional asymptotics, we derive higher-order coverage expansions and analyze the stochastic approximation step used in the implemented algorithm. For a fixed nominal level, the root of the bootstrap coverage equation is consistent under a uniform coverage…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-06-24","reactions":{}}