Posterior predictive check: CDF comparison — pp_check

Compares observed data to the posterior predictive distribution using cumulative distribution functions.

Usage

pp_check_cdf(
  fit,
  type = c("individual", "summary"),
  study_id = NULL,
  prob = 0.9
)

fit: A meta_did_fit object with method = "sample".
type: Character string: "individual" (default) for within-study outcome distributions, or "summary" for the distribution of study-level treatment effects.
study_id: Character vector of study IDs to include. NULL (default) includes all studies of the relevant type.
prob: Width of the credible ribbon. Default 0.9.

A ggplot2 object.

Two modes are available via the type argument:

"individual": For each individual-level study, compares the empirical CDF of observed outcomes to the posterior predictive CDF (with uncertainty ribbon), faceted by study and group/time cell. Requires individual-level data in the fit.
"summary": Compares the empirical CDF of observed study-level treatment effects to the posterior predictive CDF of replicated study effects. A single-panel plot that provides an aggregate calibration check across all studies.