The R package bgms provides tools for Bayesian analysis of
graphical models describing networks of variables. The package uses Markov
chain Monte Carlo methods combined with a pseudolikelihood approach to
estimate the posterior distribution of model parameters.
Gibbs variable selection GeorgeMcCulloch_1993bgms is used to model the underlying network structure of the graphical model. By imposing a discrete spike and slab prior on the pairwise interactions, it is possible to shrink the interactions to exactly zero. The Gibbs sampler embeds a Metropolis approach for mixtures of mutually singular distributions GottardoRaftery_2008bgms to account for the discontinuity at zero. The goal is to provide these tools for Markov Random Field (MRF) models for a wide range of variable types in the bgms package, and it currently provides them for analyzing networks of binary and/or ordinal variables MarsmanHaslbeck_2023_OrdinalMRFbgms.
While the goal is to provide the above tools for Markov Random Field (MRF) models for a wide range of variable types in the bgms package, it currently provides tools for analyzing networks of binary and/or ordinal variables MarsmanHaslbeck_2023_OrdinalMRFbgms.
MRFs are a special class of graphical models whose graph structure reflects the conditional associations between their variables, making them useful for testing for conditional independence or dependence. For example, the inclusion Bayes factor tests for conditional independence or dependence of a pair of variables in the network by comparing the predictive adequacy of models that include the edge between these variables and models that exclude the edge. HuthEtAl_2023_intro,SekulovskiEtAl_2023bgms.
The bgms package offers several tools for analyzing the structure of the MRF:
Simulate response data from the MRF using the Gibbs sampler.
Simulate mrfSampler.
Estimate the posterior distribution of the MRF's parameters and possibly its network structure using Gibbs variable selection.
Bayesian estimation or Bayesian edge selection with bgm.
Maintainer: Maarten Marsman m.marsman@uva.nl (ORCID)
Other contributors:
Useful links: