Rao-Blackwellized e-variables
Summary We show that for any concave utility, the expected utility of an e-variable can only increase after conditioning on a sufficient statistic, echoing the classic Rao-Blackwell theorem. The simplest form of the result has a straightforward proof using Jensen’s inequality, but we prove more sophisticated versions. We provide several applications of this insight, including a derivation of the log-optimal e-variable for testing the variance in a linear regression model.
Institutions
- Centrum Wiskunde & Informatica (NL)
- Carnegie Mellon University (US)
- Stanford University (US)
Publication Details
- Journal
- Biometrika
- Published
- 2026-10-15
- DOI
- https://doi.org/10.1093/biomet/asag061
- Primary Topic
- Statistical Methods and Inference
- Type
- article
- Field-Weighted Citation Impact
- 0.00