gbggm: Generalized Bayesian Gaussian Graphical Models

This package includes some implementations of Bayesian Gaussian Graphical Models based on the generalized approach proposed by Franco et al. (under review). Currently, there are 8 models implemented:

• Priors with one parameter for regularization (i.e., models that control the regularization with a single parameter): "normal", "laplace", "logistic", "cauchy", and "hypersec". Respectively, these values set a normal, laplace, logistic, Cauchy, or hyperbolic secant as prior distributions;
• Priors with two parameters for regularization (i.e., models that control the regularization with two parameters: a "regularization" parameter per se and an "heavy-tailedness" parameter, which may be useful when there are "outlier" correlations): "t", "lomax", and "NEG". Respectively, these values set a t, double lomax, or normal-exponential-gamma as prior distributions.

The user can also decide if they will estimate a sparse or non-sparse network. The estimation of the BGGMs is done with the bggm function.

Installation

Using the 'remotes' package:

install.packages("remotes")
remotes::install_github("vthorrf/gbggm")