Actually I think the professor is right -- that should be your initial inference if you expect it to pick up a difference and it doesn't. For example, if I mak

Actually I think the professor is right -- that should be your initial inference if you expect it to pick up a difference and it doesn't. For example, if I make a test to tell me which item weighs more between two, and then I weigh myself and my refrigerator and it says I weigh more, then I should look at the test first. Now it may turn out that the test was right, and my fridge actually is a new lightweight fridge, but your expectation is the default hypothesis. In fact the way you typically calibrate tests is to verify that it signals differences that you very strongly expect.