95% Confidence Intervals Really Cover Only 10-25% of the Time, Statistician Notes

RexDouglass · x · 2026-09-11

Rex Douglass rephrases an observation from Robin Hanson: confidence intervals implying 95% coverage of a population estimate really have only 10-25% coverage, assuming replications are additional draws from the same population. Likewise, a finding with p=0.05 has an expected replication probability of just 0.10 to 0.25 across fields — a concise reminder of how far nominal significance falls short of real reproducibility.

Related event: Researcher Claims 95% Confidence Intervals Cover Only 10-25% in Practice(2 posts)→

Original post →

More from Research

Research channel →