Articles

Getting started

These vignettes demonstrate how to use the loo package to perform approximate leave-one-out cross-validation or exact K-fold cross-validation for Bayesian models fit using MCMC, compare models on estimated predictive performance on new data, and weight models for averaging predictive distributions.

Using the loo package (version >= 2.0.0)

Bayesian Stacking and Pseudo-BMA weights using the loo package

Writing Stan programs for use with the loo package

Holdout validation and K-fold cross-validation of Stan programs with the loo package

Additional topics

These vignettes demonstrate how to use the loo package for more complicated scenarios including models with non-factorized likelihoods, forecasting models, models fit to very large datasets, and more.

Leave-one-out cross-validation for non-factorized models

Approximate leave-future-out cross-validation for Bayesian time series models

Using Leave-one-out cross-validation for large data

Avoiding model refits in leave-one-out cross-validation with moment matching

Mixture IS leave-one-out cross-validation for high-dimensional Bayesian models

Migration

Guides for updating code when APIs change.

Migration guide

Predictive performance measures

Reference and workflow articles for the *_pred_measure() API (website only; not included in the CRAN package).

Predictive model performance: Predictive schemes with `pred_measure()`

Predictive model performance: Overview of measures

Model comparison: Explanation of `model_compare()`

Frequently asked questions

Cross-validation FAQ

Answers to frequently asked questions about cross-validation and the loo package (links to external site).