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.