gp3bayes treats group-level estimates as posterior
quantities to inspect, not as automatic rankings of participants or
items.
effects <- group_effect_table(fit)
components <- variance_component_table(fit)
plot_group_effects(effects)
plot_variance_components(components)participant_ppc <- grouped_prediction_check(
fit,
group = "participant_id",
ndraws = 1000
)
as.data.frame(participant_ppc)
plot_grouped_prediction_check(participant_ppc)The check compares observed group summaries with their posterior predictive distribution. No group is automatically excluded.
uncertainty <- prediction_uncertainty_decomposition(
fit,
include_group_effects = FALSE,
ndraws = 1000
)
as.data.frame(uncertainty)
plot_uncertainty_decomposition(uncertainty)The expected-response component and remaining predictive component are Monte Carlo variance summaries under the fitted model. They should not be interpreted as a causal variance decomposition.