Beyond p<0.05: what should we teach about hypothesis testing?
Date:
The realization that many scientific studies cannot be replicated has led to calls to retire statistical significance and a ban on P-values. For many years, our Statistics classrooms have too often incorporated a procedural approach to hypothesis testing, perhaps enhanced with some caveats about what P-values are not and what we should not do. But greater consideration of the development of broader statistical thinking, including how we ask questions, design studies, and collect data, may result in better practice and better understanding of what we can conclude from scientific studies. In this workshop we will discuss how we can engage students in considerations of the scientific and statistical issues that lead to appropriate conclusions and a deeper understanding of what P-values are, starting in students? first course in Statistics. We will work through some classroom-ready examples that illustrate problems with reproducibility, discuss possible reasons, and explore simulations to develop a deeper understanding of statistical testing, including the implications of small samples and p-hacking.