Demonstrating causation through changes in variables