Skip to contents

Posterior predictive checks

Source: R/testing_through_time.R
ppc.Rd

Generate posterior predictive checks (PPCs) from a fitted Bayesian time-resolved GAMM stored in a clusters_results object. PPCs can be produced either at the group level or separately for each participant.

Usage

ppc(
  object,
  ppc_type = c("group", "participant"),
  ndraws = 500,
  group_var = NULL,
  cores = 4,
  ...
)

Arguments

object

A clusters_results object containing a fitted brmsfit model in object$model.

ppc_type

Character string specifying the type of PPC to generate. Either "group" (default) for group-level PPCs (ignoring participant identity) or "participant" for participant-wise PPCs.

ndraws

Integer specifying the number of posterior draws to use for the PPC. Defaults to 500.

group_var

Optional character; name of the grouping variable to use for grouped PPCs at the group level. If NULL (default), the function uses "predictor" when present in model$data and binary (two levels).

cores

Numeric; number of parallel cores to use (only used when ppc_type = "participant").

...

Currently unused. Included for future extensions.

Value

A ggplot object visualising the posterior predictive check. The plot is printed to the active graphics device and also returned invisibly.

Details

At the group level, predictions are obtained by simulating from the posterior using posterior_predict with re_formula = NA, after collapsing the original data across participants (by time). At the participant level, PPCs are generated using pp_check with grouped ribbons.

  • Group-level PPCs are computed by averaging numeric variables across participants at each time point, and simulating posterior predictive draws with random effects excluded (re_formula = NA). This provides a marginal, population-level posterior predictive check.

  • Participant-level PPCs are computed using grouped ribbon plots, showing posterior predictive distributions separately for each participant.

The returned object is a ggplot2 object produced by ppc_ribbon or pp_check, depending on the selected ppc_type.

See also

pp_check, posterior_predict, ppc_ribbon

Examples

if (FALSE) { # \dontrun{
# Group-level PPC
ppc(object = res, ppc_type = "group")

# Participant-level PPC
ppc(object = res, ppc_type = "participant")
} # }