tidyposterior

The tidyposterior package performs Bayesian post hoc analysis of resampling results to compare the performance of different models. It works with cross-validation or other resampling methods to make statistically rigorous comparisons between models without needing a separate test set.

The package uses Bayesian generalized linear models to analyze paired resampling statistics, treating them as posterior distributions that can be compared directly. This approach accounts for the correlation structure in resampled data and provides probabilistic statements about model differences. It integrates with the tidymodels ecosystem but can also work with any resampling results stored in a data frame.