kfold-ubmsFit-method: K-fold Cross-validation of a ubmsFit Model

In kenkellner/ubms: Bayesian Models for Data from Unmarked Animals using 'Stan'

K-fold Cross-validation of a ubmsFit Model

Description

Randomly partition data into K subsets of equal size (by site). Re-fit the model K times, each time leaving out one of the subsets. Calculate the log-likelihood for each of the sites that was left out. This function is an alternative to loo (leave-one-out cross validation).

Usage

## S4 method for signature 'ubmsFit'
kfold(x, K = 10, folds = NULL, quiet = FALSE, ...)

Arguments

x	A ubmsFit model
K	Number of folds into which the data will be partitioned
folds	An optional vector with length equal to the number of sites in the data and containing integers from 1 to K, to manually assign sites to folds. You should use this if you plan to compare multiple models, since the folds for each model should be identical. You can use loo::kfold_split_random to generate this vector
quiet	If TRUE, suppress progress bar
...	Currently ignored

Value

An object of class elpd_generic that is compatible with loo::loo_compare

kenkellner/ubms documentation built on March 1, 2025, 7:02 a.m.