Introductory Statistical Pedagogy Should Be Reformed: Transitioning from a Hypothesis Testing to Modeling Framework

The standard undergraduate statistics curriculum, which focuses on testing hypotheses, was first developed a century ago when hand-calculations dominated statistical analysis. While graduate training has advanced somewhat, the undergraduate curriculum has remained relatively unchanged for the last 50-100 years. This situation is particularly troubling when one considers how it fails to prepare students for advanced statistical methods, perpetuates misunderstandings, and masks the underlying logic of statistics. It also is highly limited in its application, leading researchers to shoehorn their analyses into a hypothesis-testing framework. An alternative to the hypothesis-testing framework is one that focuses on building, evaluating, and comparing models. In this paper, we argue that a modeling framework (and an extension of modeling called ‘model comparisons’) highlights the underlying logic of statistics, deepens understanding of the language of statistics, prepares students for advanced procedures, and promotes sound statistical reasoning. We argue for a pedagogical shift away from the hypothesis testing status quo and toward a modeling framework, all while discussing obstacles to such a change and how such obstacles might be (fairly) easily overcome.