Bayesian Priors Explained | Prior Modelling & Prior Predictive Checks

How should we choose a prior distribution in Bayesian statistics — and how can we check whether that prior actually makes sense? In this video, we explore prior modelling and prior predictive checks, two important parts of the Bayesian modelling workflow. We look at how prior distributions encode assumptions about unknown parameters, how different choices of priors influence a model, and how prior predictive simulation can reveal whether those assumptions lead to reasonable outcomes before we even look at the observed data. Topics include: • What a prior distribution represents • How to think about prior information • Choosing prior distributions • Informative and weakly informative priors • The implications of different prior choices • Prior predictive distributions • Prior predictive simulation • Prior predictive checks • Detecting unrealistic assumptions before fitting a model • Using domain knowledge to construct sensible priors The basic idea behind a prior predictive check is: Choose priors → Simulate parameters → Simulate data → Ask whether the simulated data are plausible Prior predictive checks allow us to understand the implications of our assumptions before conditioning on the observed data. This video is part of my Bayesian Statistics Explained series, where we build Bayesian modelling concepts step by step. Subscribe for more intuitive explanations of statistics, probability, Bayesian modelling, and data science.