rCausalMGM: Scalable Causal Discovery and Model Selection on Mixed Datasets with 'rCausalMGM'

Scalable methods for learning causal graphical models from mixed data, including continuous, discrete, and censored variables. The package implements CausalMGM, which combines a convex, score-based approach for learning an initial moralized graph with a producer-consumer scheme that enables efficient parallel conditional independence testing in constraint-based causal discovery algorithms. The implementation supports high-dimensional datasets and provides individual access to core components of the workflow, including MGM and the PC-Stable and FCI-Stable causal discovery algorithms. To support practical applications, the package includes multiple model selection strategies, including information criteria based on likelihood and model complexity, cross-validation for out-of-sample likelihood estimation, and stability-based approaches that assess graph robustness across subsamples.

Version:	1.0
Imports:	Rcpp (≥ 1.0.3), survival
LinkingTo:	BH, Rcpp, RcppArmadillo, RcppThread
Suggests:	Rgraphviz, graph
Published:	2026-03-03
DOI:	10.32614/CRAN.package.rCausalMGM
Author:	Tyler C Lovelace [aut], Max Dudek [aut], Jack Fiore [aut], Panayiotis V Benos [aut, cre]
Maintainer:	Panayiotis V Benos <pbenos at ufl.edu>
License:	GPL-3
NeedsCompilation:	yes
Materials:	README
CRAN checks:	rCausalMGM results [issues need fixing before 2026-03-17]

Documentation:

Reference manual:

rCausalMGM.html , rCausalMGM.pdf

Downloads:

Package source:	rCausalMGM_1.0.tar.gz
Windows binaries:	r-devel: rCausalMGM_1.0.zip, r-release: rCausalMGM_1.0.zip, r-oldrel: rCausalMGM_1.0.zip
macOS binaries:	r-release (arm64): rCausalMGM_1.0.tgz, r-oldrel (arm64): rCausalMGM_1.0.tgz, r-release (x86_64): rCausalMGM_1.0.tgz, r-oldrel (x86_64): rCausalMGM_1.0.tgz

Linking:

Please use the canonical form https://CRAN.R-project.org/package=rCausalMGM to link to this page.