Test statistics and p-values

Abstract

We point out that the traditional notion of test statistic is too narrow, and we propose a natural generalization that is arguably maximal. The study is restricted to simple statistical hypotheses.

arXiv comment22 pages