scmix: Bayesian Model-Based Clustering with Sparse Conditional Mixture
Models
Fits Bayesian sparse conditional (Gaussian) mixture models for
model-based clustering. Each mixture component factorizes into a chain
of univariate polynomial regressions with per-component, per-equation
Bayesian variable selection under a centered Zellner g-prior; the
number of clusters is selected within a single run via an overfitted
sparse mixture (Dirichlet concentration 1/K). The blocked Gibbs sampler
draws the selection sets exactly by enumeration (or by validated
single-flip Metropolis-Hastings in higher dimension), is provably
well-posed under a documented proper fallback prior, and reports a
label-invariant consensus partition (Dahl's least-squares criterion).
Companion package to Dong, Liao, and Lee (2026), "Replacing three nested searches with one sweep:
a Bayesian treatment of sparse conditional mixture clustering". Multiple-imputation
functionality for the same engine is also exposed.
Documentation:
Downloads:
Linking:
Please use the canonical form
https://CRAN.R-project.org/package=scmix
to link to this page.