Rao-Blackwellized e-variables
By: Dante de Roos , Ben Chugg , Peter Grünwald and more
We show that for any concave utility, the expected utility of an e-variable can only increase after conditioning on a sufficient statistic. The simplest form of the result has an extremely straightforward proof, which follows from a single application of Jensen's inequality. Similar statements hold for compound e-variables, asymptotic e-variables, and e-processes. These results echo the Rao-Blackwell theorem, which states that the expected squared error of an estimator can only decrease after conditioning on a sufficient statistic. We provide several applications of this insight, including a simplified derivation of the log-optimal e-variable for linear regression with known variance.
Similar Papers
Testing hypotheses generated by constraints
Statistics Theory
Finds best ways to make smart decisions.
Testing maximum entropy models with e-values
Methodology
Helps scientists test ideas with more data.
Universal Log-Optimality for General Classes of e-processes and Sequential Hypothesis Tests
Statistics Theory
Helps computers make smart guesses faster.