Are p ‐values under attack? Contribution to the discussion of ‘A critical evaluation of the current “ p ‐value controversy” ’

It is a pleasure to congratulate Prof.Wellek on his intriguing and thought-provoking paper. He reminds us of some instructive themes on this topic, and I must admit that I found little with which to quibble. My take-away message from the exposition is that while some high-visibility sources have called into question use of p-values in modern, data-rich, scientific discourse, their complaints may be overblown: the p-value is as indispensable (Prof. Wellek’s term) as ever in contemporary medical applications and in associated areas such as regulatory affairs. The issue appears to be that (i) recalcitrant objectors have argued against p-values—and null hypothesis based testing in general—as they are applied in far too-automated a fashion, and that (ii) in the end p-values provide little information about the actual experimental hypotheses under study. In reply, the American StatisticalAssociation issued a Statement in support of “statistical significance and p-values” (Wasserstein and Lazar, 2016), the useful guiding principles from which are aptly summarized, and in a few cases criticized, in Prof. Wellek’s Discussion Section. In fact, from my reading I found the agreement rather substantial between the ASA’s published principles and Prof. Wellek’s critique. I think all players would agree with Prof. Wellek’s convincing conclusion (end of his Sec. 3.3) that