B.5:8 - Common Anti-Patterns and How to Avoid Them

Recognizable failure	Repair
Repeatedly improve the fit of a model that omits the intended quantity.	Test the representation against a small action-changing counterexample; recover the missing distinction or qualify a sufficient bound.
Treat exploratory regularities as if they were independently predicted and tested.	Preserve their origin and use a design and inference that account for the selection and dependence.
Produce an argument whose receiver cannot identify what the crucial step establishes.	Work backward from the needed conclusion and recover the decisive transition and its premises. If a premise changes, follow its effect through the argument.
Generate more answers after the important question has changed.	State the changed formulation and which prior results still answer it before selecting further production.