Skip to contents

This function computes the Pareto-smoothed importance sampling leave-one-out information criterion (PSIS-LOO) for an object of class lmBayes or spmBayes. The PSIS-LOO is converted into deviance scale so that it is comparable with other information criteria like AIC and BIC.

Usage

psis.loo(object, comp.r_eff = TRUE)

Arguments

object

An object of class 'lmBayes' or 'spmBayes'.

comp.r_eff

Logical. Whether the function should estimate the relative effective sample size for the likelihood of each observation. If FALSE, all the relative effective sample sizes are set to one. Default is TRUE.

Value

The estimated PSIS-LOO.

Author

Santiago Marin