Semantics of causal DAG models and the identification of direct and indirect effects
James M. Robins · 2003
Abstract Directed acyclic graphs (DAGs) are commonly used to represent causal models. The article by Dawid posits a causal model that is closely related to the model of Spirtes et al. (1993) and the model of Pearl (1993b). In this discussion I will compare and contrast the semantics of DAGs representing the Spirtes et al. model with that of DAGs representing the non-parametric structural equation (NPSE) model of Pearl (1995a) and the finest fully randomized causally interpreted structured tree graph (FRCISTG) model of Robins (1986).