Single-Page Model Diagnostics for bayesma Objects

Produces a single-page diagnostic summary for a Bayesian meta-analysis model fitted via cmdstanr. Combines visual diagnostics (trace plots, posterior predictive check, Rhat, ESS) with a tabular summary of MCMC health indicators. All output is arranged on a single page using patchwork.

Usage

diagnostics(object, ...)

# S3 method for class 'bayesma'
diagnostics(object, pars = NULL, ndraws = 100, ...)

Arguments

object

A bayesma object containing a CmdStanMCMC fit in object$fit.

...

Additional arguments (currently unused).

pars

Character vector of parameter names for trace/ACF plots. If NULL, sensible defaults are chosen based on the model stage.

ndraws

Number of posterior draws for the posterior predictive check. Default is 100.

Value

Invisibly returns a list of diagnostic values. Prints a composite patchwork plot as a side effect.

Details

The diagnostic page contains six panels:

Top-left

Trace plots for key parameters.

Top-right

Posterior predictive check.

Middle-left

Rhat values for all parameters.

Middle-right

Effective sample size ratios for all parameters.

Bottom-left

Autocorrelation function for key parameters.

Bottom-right

MCMC diagnostics summary table.

Examples

if (FALSE) { # \dontrun{
fit <- bayesma(data, likelihood = "binomial", ...)
diagnostics(fit)
} # }