1 comment

[ 3.8 ms ] story [ 15.3 ms ] thread
The correct (admissible) decision rule for testing a hypothesis minimizes the Bayes risk, the integrated loss over the posterior distribution.

The authors don't really convey what an e-value is, or why they don't use an admissible decision rule. I'm skeptical of this research. I don't think there is any more room for basic research in this area as the book is already closed...