Testing Cumulative Advertising Effects: A Comment on Methodology