From frequentist problems towards Bayesian solutions

Abstract

The social sciences are undergoing uncertain and exciting days. Perhaps unwillingly, Daryl Bem’s 2011 paper (where he incredibly makes the case for ESP) set off a ‘replication crisis’ that shook the foundations of Psychology as a science. Dodgy methodological habits, in combination with more or less subtle questionable research practices (e.g., John et al., 2012), was the perfect recipe for a long announced disaster (Paul Meehl had already issued warnings more than 50 years ago!). Almost 10 years after Bem’s milestone paper, it has become clear that some of the problems at the base of this crisis concern the misuse of basic statistical tools. Confidence intervals, p-values, and null hypothesis significance testing are highly prevalent in scientific practice, but are also too often misued and misunderstood (e.g., Greenland et al., 2016). And importantly, they may not deliver what practitioners expect or hope, even if correctly used. It is in this very special framework that Bayesian statistics, an inferencial paradigm that has been long waiting to make its mark in Psychology, suddenly gained traction and visibility. Bayesian statistics offers a sound inferential platform through which a rational actor, in light of the observed data, logically updates his or her beliefs about a model. It is therefore of value to learn more about this paradigm. In this workshop, I will start by colouring a picture of the current state of affairs of Psychology as a science, keeping a keen eye on problems related to statistics. Afterwards, I will offer a gentle introduction to Bayesian statistics: What it is, what it requires, what it offers. I will illustrate the use of Bayesian statistics almost exclusively via the intuitive, point-and-click, JASP software https://jasp-stats.org. My goal is to showcase very simple examples that hopefully will trigger the participants’ interest for this exciting alternative way of performing data analysis.