Still, it is an error to argue in front of your data. You find yourself insensibly twisting them round to fit your theories.