1: \begin{abstract}

2:   We study worst-case growth-rate optimal (GROW) $\E$-statistics for hypothesis

3:   testing between two group models. If the underlying group $G$ acts freely on

4:   the observation space, there exists a maximally invariant statistic of the

5:   data. We show that among all $\E$-statistics, invariant or not, the likelihood

6:   ratio of the maximally invariant is GROW and that an anytime valid test can be

7:   based on this likelihood ratio. By virtue of a representation theorem of

8:   Wijsman, it is equivalent to a Bayes factor with a right Haar prior on $G$.

9:   Such Bayes factors are known to have good frequentist and Bayesian properties.

10:   We show that reductions through sufficiency and invariance can be made in

11:   tandem without affecting optimality. A crucial assumption on the group $G$ is

12:   its amenability, a well-known group-theoretical condition, 

13: %PETER 10/8 ADDED (PLEASE CHECK)  

14:   which holds for general scale- and location families as well as finite-dimensional linear regression. 

15:   %

16: \end{abstract}

17: