Biometrika Oct 15, 2026

Rao-Blackwellized e-variables

Authors
D De Roos B Chugg P Grünwald A Ramdas
View Full Paper
Paper Information
  • Journal:
    Biometrika
  • DOI:
    10.1093/biomet/asag061
  • Published:
    October 15, 2026
  • Added to Tracker:
    Oct 16, 2026
Abstract

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.

Author Details
D De Roos
Author
B Chugg
Author
P Grünwald
Author
A Ramdas
Author
Citation Information
APA Format
D De Roos , B Chugg , P Grünwald & A Ramdas (2026) . Rao-Blackwellized e-variables. Biometrika , 10.1093/biomet/asag061.
BibTeX Format
@article{paper1747,
  title = { Rao-Blackwellized e-variables },
  author = { D De Roos and B Chugg and P Grünwald and A Ramdas },
  journal = { Biometrika },
  year = { 2026 },
  doi = { 10.1093/biomet/asag061 },
  url = { https://doi.org/10.1093/biomet/asag061 }
}
Back to Papers