You make a good point, but obviously different datasets and scenarios require different methods, and the detail of any approach is less important than starting out with valid assumptions.