Discussion: Teaching hypothesis tests: Time for significant change?

Abstract

This is an opportune time to ask whether there should be important changes in the statistical syllabus for today's medical students, and in particular with regard to hypothesis tests, given the justified trend for preferring the confidence interval approach to inference. In discussing Jonathan Sterne's paper, I focus on teaching aspects, rather than the philosophical or historical elements, beginning with broader questions concerning overall course aims and objectives. To raise statistical awareness among future doctors, I argue that precisely what we teach is actually less important than how we teach. I discuss implications that only some medical students go on to become producers, as opposed to consumers, of research. I conclude that, for the time being at least, p-values are here to stay and hence still need to be taught but only alongside careful description of their various limitations. Copyright © 2002 John Wiley & Sons, Ltd.