Data Analysis: a (not so) silent revolution

Abstract

Good statistical practice is an essential component of good science. Unfortunately, in too many cases the analysis of experiments follows recipes taught already in introductory classes, sometimes even accompanied by misunderstandings about properties of statistical tests. One of these methods, significance tests based on p-values, was the widely accepted standard. So the warning by the American Statistical Association (ASA) against the use of p-values was thus a huge surprise to many. What were the reasons underlying this unique step?