And I would bet in a large subset of those cases it's because the person implementing those tests and the person interpreting the results aren't the same. And that neither understands basic statistical hypothesis testing.