Package index • loo

Package description, glossary, and included data sets

loo-package: Efficient LOO-CV and WAIC for Bayesian models
loo-glossary: LOO package glossary
loo-datasets Kline milk voice voice_loo: Datasets for loo examples and vignettes

Approximate LOO-CV

Approximate LOO-CV, Pareto smoothed importance sampling (PSIS), and diagnostics.

loo() loo_i() is.loo() is.psis_loo(): Efficient approximate leave-one-out cross-validation (LOO)
loo_subsample(): Efficient approximate LOO-CV using subsampling
loo_approximate_posterior(): Efficient approximate leave-one-out cross-validation (LOO) for posterior approximations
loo_moment_match(): Moment matching for efficient approximate leave-one-out cross-validation (LOO)
loo_moment_match_split(): Split moment matching for efficient approximate leave-one-out cross-validation (LOO)
E_loo(): Compute weighted expectations
psis() is.psis() is.sis() is.tis(): Pareto smoothed importance sampling (PSIS)
ap_psis(): Pareto smoothed importance sampling (PSIS) using approximate posteriors
tis(): Truncated importance sampling (TIS)
sis(): Standard importance sampling (SIS)
importance_sampling(): A parent class for different importance sampling methods.
weights(<importance_sampling>): Extract importance sampling weights
pareto_k_table() pareto_k_ids() pareto_k_values() pareto_k_influence_values() psis_n_eff_values() mcse_loo() plot(<psis_loo>) plot(<psis>): Diagnostics for Pareto smoothed importance sampling (PSIS)

Model comparison weighting/averaging

Functions for comparing models and computing model weights via stacking of predictive distributions or pseudo-BMA weighting.

print(<compare.loo>) model_compare(): Model comparison
loo_compare(): Model comparison (deprecated)
loo_model_weights() stacking_weights() pseudobma_weights(): Model averaging/weighting via stacking or pseudo-BMA weighting

Helper functions for K-fold CV

kfold_split_random() kfold_split_stratified() kfold_split_grouped(): Helper functions for K-fold cross-validation
kfold() is.kfold(): Generic function for K-fold cross-validation for developers

Predictive performance measures

Unified API for predictive performance measures (density scores, scoring rules, and point-prediction metrics) with in-sample, LOO, K-fold, and holdout evaluation modes.

insample_pred_measure(): In-sample predictive performance measures
loo_pred_measure(): PSIS-LOO predictive performance measures
kfold_pred_measure(): K-fold cross-validation predictive performance measures
test_pred_measure(): Holdout predictive performance measures
pred_measure(): Add predictive performance measures to an existing result
custom_measure(): Define a custom predictive measure
supported_measures_list: Supported predictive measure names
measure_acc(): Classification accuracy (acc)
measure_bacc(): Balanced classification accuracy (bacc)
measure_brier(): Brier score (brier)
measure_elpd(): Expected log pointwise predictive density (elpd)
measure_ic(): Information Criteria (ic)
measure_mae(): Mean absolute error (mae)
measure_mlpd(): Mean log pointwise predictive density (mlpd)
measure_mse(): Mean squared error (mse)
measure_r2(): Predictive R-squared (r2)
measure_rmse(): Root mean squared error (rmse)
measure_rps(): Ranked Probability Score (RPS, SRPS, CRPS, SCRPS)
measure_srps(): Scaled Ranked Probability Score (SRPS, SCRPS)
ptw_log_pred_density(): Pointwise log predictive density (lppd_i)

Other functions

waic() is.waic(): Widely applicable information criterion (WAIC)
extract_log_lik(): Extract pointwise log-likelihood from a Stan model
pointwise(): Convenience function for extracting pointwise estimates
relative_eff(): Convenience function for computing relative efficiencies
find_model_names(): Find the model names associated with "loo" objects
gpdfit(): Estimate parameters of the Generalized Pareto distribution
example_loglik_array() example_loglik_matrix(): Objects to use in examples and tests
print(<loo>) print(<waic>) print(<psis_loo>) print(<importance_sampling_loo>) print(<psis_loo_ap>) print(<psis>) print(<importance_sampling>) print(<kfold>): Print methods
nobs(<psis_loo_ss>): The number of observations in a psis_loo_ss object.
obs_idx(): Get observation indices used in subsampling
update(<psis_loo_ss>): Update psis_loo_ss objects

Deprecated functions

elpd(): Generic (expected) log-predictive density (deprecated)
crps() scrps() loo_crps() loo_scrps(): Continuously ranked probability score (deprecated)
loo_predictive_metric(): Estimate leave-one-out predictive performance (deprecated)
compare(): Model comparison (deprecated, old version)
psislw(): Pareto smoothed importance sampling (deprecated, old version)