In this talk, we will show how to use simulations in Stata to explore to what extent and under what circumstances a test is problematic. We will illustrate this for a set of tests of the parallel regression assumption in ordered logit and probit models: the Brant, likelihood ratio, Wald, score,...