Why, When & How to use Bayes Factors

Abstract

A great franchise comes with a great rivalry: Marvel has Iron Man and Captain America; physics has General Relativity and Quantum Physics; and Bayesian stats has Posterior Estimation and… Bayes Factors!

A few months ago, I had the pleasure of hosting EJ Wagenmakers, to talk about these topics. This time, I’m talking with Jorge Tendeiro, who has a different perspective on Null Hypothesis Testing in the Bayesian framework, and its relationship with generative models and posterior estimation.

But this is not your classic, click-baity podcast, and I’m not interested in pitching people against each other. Instead, you’ll hear Jorge talk about the other perspective fairly, before even giving his take on the topic. Jorge will also tell us about the difficulty of arguing through papers, and all the nuances you lose compared to casual discussions.

But who is Jorge Tendeiro? He is a professor at Hiroshima University in Japan, and he was recommended to me by Pablo Bernabeu, a listener of this very podcast.

Before moving to Japan, Jorge studied math and applied stats at the University of Porto, and did his PhD in the Netherlands. He focuses on item response theory (specifically person fit analysis), and, of course, Bayesian statistics, mostly Bayes factors.

He’s also passionate about privacy issues in the 21st century, an avid Linux user since 2006, and is trying to get the hang of the Japanese language.