Pair-copula constructions for non-Gaussian Bayesian networks
Alexander Bauer · 2013
We propose a new multivariate statistical model that permits non-Gaussian distributions as well as the inclusion of conditional independence assumptions specified by a directed acyclic graph. This combination of features is achieved by using pair-copula constructions. We provide routines for random sampling and likelihood inference, and investigate model selection. Structure estimation is facilitated using a version of the PC algorithm that is based on a novel test for conditional independence.