Growth-Rate Optimal Gaussian Mixed $E$-Processes for One-Way ANOVA
Alexander Ly ⋅ Udo Boehm ⋅ Wouter Koolen
Abstract
We derive growth-rate optimal Gaussian mixed test martingales for the $K$-sample problem that enables anytime-valid inference under adaptive treatment allocation. This approach is reference group invariant, growth-rate optimal, and allows for simultaneous anytime-valid confidence sets for the population mean differences.
Chat is not available.
Successful Page Load