Graphs for margins of Bayesian networks

TL;DR

This paper introduces mDAGs, a new class of hyper-graphs, to accurately represent the complex marginal models of Bayesian networks with latent variables, surpassing traditional mixed graphs.

Contribution

The paper proposes mDAGs and a latent projection method, enabling precise representation of all marginal models derived from DAGs with latent variables.

Findings

mDAGs can represent all marginal models of DAGs with latent variables

Graphical criteria for when two marginal models are equivalent

mDAGs capture causal structure under interventions

Abstract

Directed acyclic graph (DAG) models, also called Bayesian networks, impose conditional independence constraints on a multivariate probability distribution, and are widely used in probabilistic reasoning, machine learning and causal inference. If latent variables are included in such a model, then the set of possible marginal distributions over the remaining (observed) variables is generally complex, and not represented by any DAG. Larger classes of mixed graphical models, which use multiple edge types, have been introduced to overcome this; however, these classes do not represent all the models which can arise as margins of DAGs. In this paper we show that this is because ordinary mixed graphs are fundamentally insufficiently rich to capture the variety of marginal models. We introduce a new class of hyper-graphs, called mDAGs, and a latent projection operation to obtain an mDAG from the margin of a DAG. We show that each distinct marginal of a DAG model is represented by at least one mDAG, and provide graphical results towards characterizing when two such marginal models are the same. Finally we show that mDAGs correctly capture the marginal structure of causally-interpreted DAGs under interventions on the observed variables.