abn: an R package for modelling multivariate data using additive Bayesian networks

This vignette describes the R package abn which provides functionality for identifying statistical dependencies in complex multivariate data using additive Bayesian network (ABN) models. This methodology is ideally suited for both univariate one response variable, and multiple explanatory variables and multivariate analysis, where in both cases all statistical dependencies between all variables in the data are sought. ABN models comprise of directed acyclic graphs (DAGs) where each node in the graph comprises a generalized linear model. Model search algorithms are used to identify those DAG structures most supported by the data. Currently implemented are models for data comprising of categorical, count and/or continuous variables. Further relevant information about abn can be found at: www.r-bayesian-networks.org.