Skip to contentSkip to filters

We use cookies for analytics.

#169 Bayesian Workflow for Beginners, Alex at Princeton

Related content

About

In a talk at Princeton, Alexandre Andorra walks through the Bayesian workflow for the mixed models psychologists fit to repeated-measures data: writing the model as a generative story, the hidden assumptions of a standard mixed model (stable latent abilities, uniform condition effects, independent trials), how lme4, brms and PyMC express the same model, prior predictive checks and why default priors such as Normal(0, 10) are risky, reading posterior distributions and credible intervals, partial pooling and shrinkage of extreme participants, and checking whether a model is well calibrated. He also advises starting with the simplest model and adding complexity one layer at a time.

Topics

Details

Show

Learning Bayesian Statistics

5 episodes on World Wide · Learning Bayesian Statistics