Why is it important to control all variables except one when studying cause-and-effect relationships?.