Actually, that is a huge problem. You have to decide on the test you are going to use before you see the data. After you've seen the data you can always come up

Actually, that is a huge problem. You have to decide on the test you are going to use before you see the data. After you've seen the data you can always come up with a specially made statistical test that will prove anything you want with any p-value you want.