Practical Bayesian model evaluation using leave-one-out cross-validation and WAIC

Leave-one-out cross-validation (LOO) and the widely applicable information criterion (WAIC) are methods for estimating pointwise out-of-sample prediction accuracy from a fitted Bayesian model using the log-likelihood evaluated at the posterior simulations of the parameter values. LOO and WAIC have v...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Statistics and computing Jg. 27; H. 5; S. 1413 - 1432
Hauptverfasser: Vehtari, Aki, Gelman, Andrew, Gabry, Jonah
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York Springer US 01.09.2017
Springer Nature B.V
Schlagworte:
Artificial Intelligence
Bayesian analysis
Errors
Importance sampling
Mathematics and Statistics
Probability and Statistics in Computer Science
Statistical inference
Statistical Theory and Methods
Statistics
Statistics and Computing/Statistics Programs
Stan
Widely applicable information criterion (WAIC)
Bayesian computation
Pareto smoothed importance sampling (PSIS)
Leave-one-out cross-validation (LOO)
fold cross-validation
ISSN:0960-3174, 1573-1375
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Leave-one-out cross-validation (LOO) and the widely applicable information criterion (WAIC) are methods for estimating pointwise out-of-sample prediction accuracy from a fitted Bayesian model using the log-likelihood evaluated at the posterior simulations of the parameter values. LOO and WAIC have various advantages over simpler estimates of predictive error such as AIC and DIC but are less used in practice because they involve additional computational steps. Here we lay out fast and stable computations for LOO and WAIC that can be performed using existing simulation draws. We introduce an efficient computation of LOO using Pareto-smoothed importance sampling (PSIS), a new procedure for regularizing importance weights. Although WAIC is asymptotically equal to LOO, we demonstrate that PSIS-LOO is more robust in the finite case with weak priors or influential observations. As a byproduct of our calculations, we also obtain approximate standard errors for estimated predictive errors and for comparison of predictive errors between two models. We implement the computations in an R package called loo and demonstrate using models fit with the Bayesian inference package Stan.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0960-3174
1573-1375
DOI:10.1007/s11222-016-9696-4