loo: Efficient Leave-One-Out Cross-Validation and WAIC for Bayesian Models

Efficient approximate leave-one-out cross-validation (LOO) for Bayesian models fit using Markov chain Monte Carlo, as described in Vehtari, Gelman, and Gabry (2017) <doi:10.1007/s11222-016-9696-4>. The approximation uses Pareto smoothed importance sampling (PSIS), a new procedure for regularizing importance weights. As a byproduct of the calculations, we also obtain approximate standard errors for estimated predictive errors and for the comparison of predictive errors between models. The package also provides methods for using stacking and other model weighting techniques to average Bayesian predictive distributions.

Version: 2.5.1
Depends: R (≥ 3.1.2)
Imports: checkmate, matrixStats (≥ 0.52), parallel, stats
Suggests: bayesplot (≥ 1.7.0), brms (≥ 2.10.0), ggplot2, graphics, knitr, rmarkdown, rstan, rstanarm (≥ 2.19.0), rstantools, spdep, testthat (≥ 2.1.0)
Published: 2022-03-24
Author: Aki Vehtari [aut], Jonah Gabry [cre, aut], Mans Magnusson [aut], Yuling Yao [aut], Paul-Christian Bürkner [aut], Topi Paananen [aut], Andrew Gelman [aut], Ben Goodrich [ctb], Juho Piironen [ctb], Bruno Nicenboim [ctb]
Maintainer: Jonah Gabry <jsg2201 at columbia.edu>
BugReports: https://github.com/stan-dev/loo/issues
License: GPL (≥ 3)
URL: https://mc-stan.org/loo/, https://discourse.mc-stan.org
NeedsCompilation: no
SystemRequirements: pandoc (>= 1.12.3), pandoc-citeproc
Citation: loo citation info
Materials: NEWS
In views: Bayesian
CRAN checks: loo results

Documentation:

Reference manual: loo.pdf
Vignettes: Holdout validation and K-fold cross-validation of Stan programs with the loo package
Using the loo package
Using Leave-one-out cross-validation for large data
Approximate leave-future-out cross-validation for Bayesian time series models
Avoiding model refits in leave-one-out cross-validation with moment matching
Leave-one-out cross-validation for non-factorized models
Bayesian Stacking and Pseudo-BMA weights
Writing Stan programs for use with the loo package

Downloads:

Package source: loo_2.5.1.tar.gz
Windows binaries: r-devel: loo_2.5.1.zip, r-release: loo_2.5.1.zip, r-oldrel: loo_2.5.1.zip
macOS binaries: r-release (arm64): loo_2.5.1.tgz, r-oldrel (arm64): loo_2.5.1.tgz, r-release (x86_64): loo_2.5.1.tgz, r-oldrel (x86_64): loo_2.5.1.tgz
Old sources: loo archive

Reverse dependencies:

Reverse depends: bistablehistory, evidence, spsurv, TriDimRegression
Reverse imports: BAMBI, BartMixVs, bayesbr, bayesdfa, bayesforecast, bayesGAM, bayesnec, beanz, blavaan, bmgarch, bmggum, bmscstan, brms, causalOT, conformalbayes, disbayes, FlexReg, glmmfields, hBayesDM, hsstan, mcmcsae, mcp, MetaStan, missingHE, MixSIAR, pcFactorStan, projpred, publipha, rater, rbioacc, rmsb, rstan, rstanarm, rtmpt, SPQR, stanette, StanMoMo, ubms, walker
Reverse suggests: bayesplot, bayesvl, bmstdr, CopulaDTA, footBayes, multinma, performance, redist, rPBK, tipsae

Linking:

Please use the canonical form https://CRAN.R-project.org/package=loo to link to this page.