Skip to contents

GHK-simulated log-likelihood of each participant's sort under each kept posterior draw (thinned to draws), the input PSIS-LOO needs. The GHK seed depends only on the draw index and the person, never on K, so log-likelihoods are common-random-number comparable across a ladder of fits.

Usage

loglik_person(fit, draws = 100, R = 64, seed = 1)

Arguments

fit

A bayesqm_fit.

draws

Posterior draws to evaluate (default 100, thinned evenly).

R

GHK replications per evaluation (default 64).

seed

Base seed for the GHK draws (default 1).

Value

A draws x N matrix of log-likelihoods.