Empirical Bayes estimation via data fission

TL;DR

This paper introduces data fission as a technique to generate synthetic replicates from single observations, enabling empirical Bayes estimation in single-replicate settings by framing it as a regression problem.

Contribution

It extends empirical Bayes methods to single-replicate data using data fission, bridging the gap with multiple-replicate approaches.

Findings

Data fission effectively creates synthetic replicates.

Empirical Bayes estimation can be formulated as a regression after data fission.

The method broadens empirical Bayes applicability to single observations.

Abstract

We demonstrate how data fission, a method for creating synthetic replicates from single observations, can be applied to empirical Bayes estimation. This extends recent work on empirical Bayes with multiple replicates to the classical single-replicate setting. The key insight is that after data fission, empirical Bayes estimation can be cast as a general regression problem.