In general, forward iterative model building is especially bad for NHST, given an assumption is a fixed model, ie, no model selection. That said, especially in the social behavior sciences, incorrect approaches are actually taught and written about in books…lol… no one seems to mind.