So, as I understand it: how do you pick a P such that, if you fail to prove there is a difference with p<P, then there is no difference? If you compute the pow

So, as I understand it: how do you pick a P such that, if you fail to prove there is a difference with p<P, then there is no difference? If you compute the power of your test, you can do this, so long as you know the expected difference of your means. Like you implied, that is not realistic. Do you do something else entirely: https://en.m.wikipedia.org/wiki/Equivalence_test https://en.m.wikipedia.org/wiki/Equivalence_test