Obtain the leave-one-out information criterion (LOO IC)

Extracts pointwise log-likelihood values and calculates Pareto-smoothed importance-sampling leave-one-out statistics.

Usage

get_loo(fit)

Arguments

fit

A sampled sbt_fit() object containing the model data and posterior draws.

Value

A psis_loo object containing LOO estimates and diagnostics.

Details

The calculation includes AF, catch LF, CPUE LF, CPUE index, aerial, troll, conventional tagging, POP, HSP, and genetic-tag components when their switches are active. Priors, process penalties, and harvest-wall contributions are not pointwise observation likelihoods and are excluded. AF, catch-LF, and CPUE-LF rows with zero input effective sample size are retained in the model data and output structures but excluded from the pointwise LOO calculation.