Trials and Tribulations of Teaching NHST in the Health Sciences

Abstract

Null hypothesis significance testing (NHST) with a critical level of significance of 5% (P<0.05) has become the cornerstone of research in the health sciences, underpinning decision making. However, considerable debate exists about its value with claims it is misused and misunderstood. It has been suggested it is because NHST and P-values are too difficult to teach, and encourage dichotomous thinking in students. Consequently, as part of statistics reform it has been proposed NHST should no longer be taught in introductory courses. However, this presentation will consider if the misuse of NHST principally results from it being taught in a mechanistic way, along with claims to knowledge in teaching and erosion of good practice. Whilst hypothesis testing has shortcomings, it is advocated it is an essential component of the undergraduate curriculum. Students’ understanding can be enhanced by providing philosophical perspectives to statistics, supplemented by overviews of Fisher’s and Neyman-Pearson’s theories. This helps the appreciation of the underlying principles of statistics based on uncertainty and probability, plus the contrast of statistical with contextual significance. Moreover, students need to appreciate when to use NHST rather than being taught it as the definitive approach of drawing inferences from data.