GRAPHS, CAUSALITY, AND STRUCTURAL EQUATION MODELS

Structural equation models (SEA ls) have dominated causal analysis in the social and behavioral sciences since the 1960s. Currently, many SE M practitioners are having difficulty articulating the causal content of SEM and an seeking foundational answers. Recent developments in the areas of graphical models and the logic of causality show potential f or alleviating such difficulties and thus, revitalizing structural equ ations as the primary language of causal modeling. This article summar izes several of these developments, including the prediction of vanish ing partial correlations, model testing, model equivalence, parametric and nonparametric identifiability, control of confounding, and covari ate selection These developments clarify the causal and statistical co mponents of SEMs and the role of SEM in the empirical sciences.