ivd() accepts arbitrary grouping IDs (character, factor, or numeric with
gaps, e.g. real school codes). IDs are recoded internally to the 1..J index
NIMBLE needs -- without reordering the data -- and the original labels are
stored on the fitted object as group_labels.summary() and plot() gained a shared labels = c("index", "original")
argument: the default keeps the compact internal cluster index in both
displays; "original" switches both to the user's own grouping IDs.print() method for fitted objects: formulas, data dimensions,
sampling setup, and a one-line convergence note (max split-Rhat, min
n_eff).pip() extractor returning the posterior inclusion probabilities as a
tidy data frame (one row per cluster x scale random effect, with the
original cluster IDs).fixef(), ranef(), and coef() extractors following the lme4/nlme
conventions (without adding a dependency on either).pip_sensitivity(): analytic prior-sensitivity for the PIPs. The
posterior odds are converted to a prior-independent Bayes factor and
re-weighted across a grid of prior inclusion probabilities -- no refit
needed. Comes with a plot() method showing per-cluster PIP trajectories
(prior-sensitive clusters highlighted; subset with clusters =).summary() now returns a structured summary.ivd object ($table,
$waic, $lppd, $pwaic, $chains) rendered by a separate print()
method; console output is unchanged.codaplot() no longer requires the user to attach coda: plot types such as
"traceplot" are now resolved in the coda namespace, so
ivd::codaplot(fit) works from a plain script (previously it failed with
object 'traceplot' not found unless library(coda) had been called).n_eff = "local" diagnostics on short or
strongly autocorrelated chains (Geyer truncation over an empty set
produced 1:Inf)..autocorrelation_fft() now actually zero-pads, computing a linear rather
than circular autocorrelation; results match stats::acf().summary() and
codaplot() for models where the number of location and scale random
effects differ.DESCRIPTION now correctly states that the spike-and-slab prior is on the
scale random effects, and cites the published paper:
Carmo, Williams & Rast (2026) doi:10.3102/10769986261426004.Initial CRAN release