Well, you're right; there could be hundreds of confounding factors, but one thing we can't do is experiment with control groups. I would think multiple regression is going to point the way.