Causal inference via ancestral graph models
Thomas Stuart Richardson, Peter Spirtes · 2003
Abstract In this article we introduce a new class of graphical models, called ancestral graph Markov models. These models are designed to provide a framework for making inferences about causal structure, and in special cases treatment effects, from background knowledge and non-randomized samples. We first provide some informal background and motivation for the models we subsequently introduce; the concepts used here will be defined more precisely in later sections.