donaldRwilliams/BGGM: Bayesian Gaussian Graphical Models

This package is described in Williams and Mulder (2019) and Williams (2018). The methods are seperated into two Bayesian approaches for inference: hypothesis testing and estimation. The former is described in Williams and Mulder (2018a), and allows for testing for the presence of edges with the Bayes factor. One-sided hypothesis testing is also possible. These methods can also provide evidence for the null hypothesis. There are extensions for confirmatory hypothesis testing in GGMs, that can include inequality or equality contraints on the partial correlation. The estimation based method are described in Williams (2018). The methods offer advantages compared to classical methods, in that a measure of uncertainty is provided for all parameters. For example, each node has a distribution for the variance explained. Measure of out-of-sample performance are also available. The model is selected with directional Bayes factors.

Getting started

Package details

MaintainerDonald R. Williams <[email protected]>
Package repositoryView on GitHub
Installation Install the latest version of this package by entering the following in R:
donaldRwilliams/BGGM documentation built on Feb. 17, 2019, 5:22 p.m.