Fragility Part I: a guide to understanding statistical power

Abstract

The aim of this paper is to close the knowledge-to-practice gap around statistical power. We demonstrate how four factors affect power: p value, effect size, sample size, and variance. This article further delves into the advantages and disadvantages of a priori versus post hoc power analyses, though we believe only understanding of the former is essential to addressing the present-day issue of reproducibility in research. Upon reading this paper, physician-scientists should have expanded their arsenal of statistical tools and have the necessary context to understand statistical fragility.