Comment on Murtaugh

Abstract

Murtaugh and I agree on an important point: abandon accept/reject declarations. That alone will go a long way to improving statistical practice. (2) Don't confuse P values or deltaAIC with binary declarations. An argument against one is not necessarily an argument against the other. (3) Be careful interpreting a P value or deltaAIC as strength of evidence. That interpretation cannot be made formal and the connection between P, deltaAIC, and evidence must be recalibrated for each new problem. (4) Plot. Check models. Plot. Check assumptions. Plot.