|
Abstract:
|
In an empirical Bayes analysis, we use data from repeated sampling to imitate inferences made by an oracle Bayesian with extensive knowledge of the data-generating distribution. Existing results provide a comprehensive characterization of when and why empirical Bayes point estimates accurately recover oracle Bayes behavior. In this paper, we develop confidence intervals that provide asymptotic frequentist coverage of empirical Bayes estimands. Our intervals include an honest assessment of bias even in situations where empirical Bayes point estimates may converge very slowly. Applied to multiple testing situations, our approach provides flexible and practical confidence statements about the local false sign rate. As an intermediate step in our approach, we develop methods for inference about linear functionals of the effect size distribution in the Bayes deconvolution problem that may be of independent interest.
|
