Thomas Lumley • 6/11/2019

Confidence intervals: not a very strong property

This technical article critiques confidence intervals as a weak statistical property, using examples like a contrived dice-roll procedure and a recent paper on small-area sampling. It explains how flexible, data-dependent interval constructions can still formally meet coverage requirements while being practically questionable, blending frequentist and Bayesian ideas.

0 comments

#data analysis #statistics #sampling