The causal manipulation of chain event graphs

TL;DR

This paper introduces chain event graphs (CEGs), a flexible graphical model that generalizes Bayesian networks for causal analysis, allowing for effective causal effect identification through graph topology.

Contribution

It presents CEGs as a new class of models that extend Bayesian networks, providing a framework for causal inference based on graph topology analysis.

Findings

CEGs generalize Bayesian networks for causal modeling.

Identifiability of causal effects can be analyzed via CEG graph topology.

Theorems analogous to the back-door criterion are established for CEGs.

Abstract

Discrete Bayesian Networks have been very successful as a framework both for inference and for expressing certain causal hypotheses. In this paper we present a class of graphical models called the chain event graph (CEG) models, that generalises the class of discrete BN models. It provides a flexible and expressive framework for representing and analysing the implications of causal hypotheses, expressed in terms of the effects of a manipulation of the generating underlying system. We prove that, as for a BN, identifiability analyses of causal effects can be performed through examining the topology of the CEG graph, leading to theorems analogous to the back-door theorem for the BN.