On comparing different tests of the same hypothesis