Eight reasons not to test for baseline group equivalence in a parallel groups pretest-posttest study