Frameworks for Credible Causal Inference in Observational Studies of Family Violence

Causal questions pervade family violence research, including how and why violence occurs, how it can be prevented, and how to support survivors. Randomization cannot always be used, but in studies of the effects of known causes, regression based observational methods (e.g., propensity score analysis) can be used. Statistical findings are associational only, however, and plausible assumptions are needed to give findings causal meaning. In this review, I suggest two criteria for plausible assumptions that help support credible causal inferences. I then describe and connect three frameworks that provide standards for specifying assumptions in causal inference: the Rubin potential outcomes framework; the Pearl directed acyclic graph framework; and the Campbell framework of validities (e.g., internal validity). These frameworks are widely accepted across the social sciences, facilitating dissemination and critique. Each has strengths and weaknesses, and in studies of the effects of known causes, they may complement each other. Utilizing the standards given to us by these frameworks to select and articulate the assumptions needed for credible causal inference will impact our understanding of family violence and violence prevention more so than studies that cannot be connected as clearly to such standards.