When developing new statistical methods, it is very useful to test them on both fake data (i.e., simulations) and real data. Testing on fake data is useful because then you know the “true” answer and can check the procedure under ideal conditions. If your method doesn’t work when the data are designed for the task, it is unlikely to work in real conditions. Fake data also enables you to test the robustness of your method when the conditions aren’t perfect – for example, try adding some nasty outliers and see if the method still …