Skip to contents

Run Integrated Workflow (Prior, MCMC, Posterior)

run_pgdcm_auto.Rd

Orchestrates the complete end-to-end Bayesian estimation pipeline for PGDCM and SEM. Computes prior predictive checks, compiles and runs Nimble MCMC, assesses convergence, and generates posterior predictive checks.

Usage

run_pgdcm_auto(
  config,
  estimation_config = list(niter = 1000, nburnin = 100, chains = 2, prior_sims = NULL,
    post_sims = NULL),
  prefix = NULL,
  threshold = 0.5,
  return_groups = FALSE,
  estimate = TRUE
)

Arguments

config

Model configuration list from build_model_config.

estimation_config

List of parameters controlling MCMC execution (niter, nburnin, chains, prior_sims, post_sims). The prior_sims and post_sims arguments control the number of simulated datasets drawn during Prior and Posterior Predictive Checking respectively. Passing NULL disables these checks entirely. Defaults to list(niter = 1000, nburnin = 100, chains = 2, prior_sims = NULL, post_sims = NULL).

prefix

Character. Descriptor prefix used for saving generated reports. Defaults to a timestamped string based on the model type.

threshold

Numeric. The mastery probability threshold to use for latent class grouping. Default is 0.5.

return_groups

Logical. If TRUE, the model groups participants into exhaustive latent classes (Caution: Scales exponentially with attributes). Default is FALSE.

estimate

Logical. If TRUE (default), MCMC estimation is performed. If FALSE, estimation is skipped (e.g., to only run prior predictive checks).

Value

A comprehensive list containing:

  • mcmc_out: The raw Nimble MCMC output list.

  • samples: The cleaned mcmc.list with structural NAs removed.

  • mapped_parameters: A data.frame of all summary statistics mapped from model parameter names to human-readable names.

  • skill_profiles: Generated by generate_summary_tables(); An I x K matrix of posterior mean mastery for each student.

  • item_parameters: Generated by generate_summary_tables(); A clean table of difficulty and discrimination parameters per item.

  • group_patterns: Generated by generate_summary_tables(); A list organizing participants into latent classes based on 0.5 mastery thresholds.

  • prior_ppc: Results from the prior predictive check (if requested).

  • post_ppc: Results from the posterior predictive check (if requested).

  • WAIC: The Watanabe-Akaike Information Criterion metric.