Compute MCMC Diagnostics for a DPMirt Fit

Returns a structured list of diagnostic information including effective sample sizes (ESS), optional chain-aware R-hat, WAIC, log-likelihood trace, timing, and DPM-specific cluster diagnostics (number of clusters, alpha posterior).

Usage

dpmirt_diagnostics(fit)

# S3 method for class 'dpmirt_diagnostics'
print(x, ...)

Arguments

fit

A dpmirt_fit object from dpmirt.

x

A dpmirt_diagnostics object.

...

Additional arguments (currently unused).

Value

A dpmirt_diagnostics S3 object containing:

ess

List of ESS vectors for items and persons.

ess_by_chain

Per-chain ESS data frames when chain metadata exists.

ess_min_items, ess_min_theta

Minimum finite item/person ESS.

rhat

R-hat vectors by parameter family when computable.

rhat_max

Maximum finite R-hat value, or NA.

waic

WAIC value (if computed).

waic_by_chain

Per-chain WAIC provenance when available.

waic_aggregation

How top-level WAIC was aggregated.

loglik_trace

Log-likelihood trace vector.

loglik_by_chain

Log-likelihood trace with chain metadata.

chain_info

Stored chain/run metadata when available.

n_clusters

Posterior cluster counts (DPM only).

n_clusters_summary

Summary of posterior cluster counts (DPM only).

alpha_summary

Alpha posterior summary (DPM only).

compilation_time, sampling_time, total_time

Timing information.

Details

Effective sample size (ESS) measures the number of effectively independent draws from the posterior. Low ESS (< 100) suggests poor mixing and the need for longer chains or different samplers. R-hat is reported only when draw_index contains at least two labeled chains with retained draws. Sequential resume segments are tracked in run_history but are not treated as independent chains. For DPM models, the cluster count trace is a first-pass visual check: stable oscillation is reassuring, while monotonic trends suggest the run has not yet mixed.

See also

dpmirt, dpmirt_compare, plot.dpmirt_fit

Other diagnostics: dpmirt_compare()

Examples

if (FALSE) { # \dontrun{
sim <- dpmirt_simulate(200, 20, model = "rasch", seed = 42)
fit <- dpmirt(sim$response, model = "rasch", prior = "dpm",
              niter = 5000, nburnin = 1000, seed = 123)
diag <- dpmirt_diagnostics(fit)
print(diag)
} # }