Posterior predictive check for bmmfit objects

For models where brms provides pp_check support, this method delegates to brms::pp_check(). For models with multinomial families (e.g., the m3 model), brms's pp_check is unavailable; this method dispatches to a model-specific visualisation instead.

Usage

# S3 method for class 'bmmfit'
pp_check(
  object,
  type = NULL,
  ndraws = NULL,
  group = NULL,
  resp_var = NULL,
  ...
)

Arguments

object

A bmmfit object returned by bmm().

type

Character. Type of pp_check. When NULL (default), resolves to "dens_overlay", or to the selected observable's default type when resp_var is specified. When group is specified, the grouped variant (e.g., "dens_overlay_grouped") is auto-selected if available. Multinomial models produce a response proportion profile regardless of the value supplied. With resp_var, type = "bars_binned" is also available: it bins a continuous statistic like a histogram, with bars for the observed number of observations per bin and points with intervals for the predicted number. It is the default for the ezdm() accuracy check.

ndraws

Integer. Number of posterior draws. Defaults to 100 for multinomial models and 10 when resp_var is specified; otherwise passed to brms::pp_check().

group

Character. Optional grouping variable for faceting. For non-multinomial models, passed to brms::pp_check(); when specified, the grouped variant of type (e.g., "dens_overlay_grouped") is auto-selected if available. For multinomial models, facets by the named predictor.

resp_var

Character. For models that declare several observables, the name of the observable to check, or "all" for a panel of all available checks built from one shared simulation. See pp_check_vars() for the options of a fitted model. The default NULL checks the primary response via brms::pp_check(), except for an ezdm(version = "4par") fit in which some cells have no usable summaries at the upper boundary: the primary response, mean_rt_upper, holds placeholders there, so NULL means resp_var = "mean_rt_upper", which leaves those cells out but takes neither newdata nor the loo_* types. For the RT models, passing negative_rt = TRUE (a brms::posterior_predict() argument) is redirected to resp_var = "signed_rt", so that observed and predicted response times are both signed by the response.

...

Additional arguments. Without resp_var, forwarded to brms::pp_check(), or for multinomial models to brms::posterior_predict() (probs, a numeric vector of length 2 with default c(0.025, 0.975), sets the credible interval). With resp_var, draw_ids and re_formula go to brms::prepare_predictions() and the rest to the bayesplot::ppc_* function. type = "bars_binned" takes breaks (bin edges that cover the observed and predicted values), prob (interval width, default 0.9) and freq (FALSE for proportions instead of counts). re_formula = NA predicts at the population level on every path.

Value

For multinomial models, for a 4-parameter ezdm() fit with placeholders in mean_rt_upper, or when resp_var is specified, a ggplot2 object (a bayesplot_grid for resp_var = "all"). For other models, the result of brms::pp_check().

Details

For multinomial models, the plot mirrors the bayesplot ppc_bars style: observed proportions are shown as bars and posterior predictive medians with credible intervals are shown as point-ranges, using the bayesplot default colour scheme and theme.

Some models describe several observables jointly (e.g. ddm(): response times and responses; ezdm(): mean RT, RT variance and accuracy), but brms only ever checks the primary response. For these models the resp_var argument selects which observable to check; pp_check_vars() lists the available checks. All selected observables are drawn from one joint posterior predictive simulation, so resp_var = "all" panels are mutually consistent.

Some observables are undefined for some cells — an ezdm() boundary has no mean response time when fewer than two responses reach it. Observations whose observed value is undefined are dropped from the check; undefined values in the simulated replicates are absorbed by dropping those posterior draws instead, so the number of observations checked does not depend on ndraws. Dropped draws are not missing at random — they are draws whose parameters made a boundary sparse — so the retained predictive is mildly conditioned; both reductions are reported with a warning. With resp_var = "all" one reduction is shared by every panel, so the panels are computed on the same observations and draws.

See also

brms::pp_check(), pp_check_vars()