🤖 AI Summary
This paper addresses the testability of unmeasured confounding in observational studies, aiming to determine whether valid causal inference is feasible. We propose the first statistically rigorous method to test the “no unmeasured confounding” assumption, achieved by formally establishing a mathematical correspondence between the potential outcomes framework and causal graph models—thereby clarifying the fundamental distinction between causal identification and conventional association-based inference. Our approach operates within linear structural equation models and leverages joint analysis of randomized controlled trial (RCT) data and observational data to calibrate statistical power and rigorously control Type I error. The key contribution is the first empirically implementable, reproducible diagnostic test for unmeasured confounding, providing practitioners with a practical tool to assess the credibility and scope of causal conclusions drawn from observational studies.
📝 Abstract
This paper clarifies a fundamental difference between causal inference and traditional statistical inference by formalizing a mathematical distinction between their respective parameters. We connect two major approaches to causal inference, the potential outcomes framework and causal structure graphs, which are typically studied separately. While the unconfoundedness assumption in the potential outcomes framework cannot be assessed from an observational dataset alone, causal structure graphs help explain when causal effects are identifiable through graphical models. We propose a statistical test to assess the unconfoundedness assumption, equivalent to the absence of unmeasured confounding, by comparing two datasets: a randomized controlled trial and an observational study. The test controls the Type I error probability, and we analyze its power under linear models. Our approach provides a practical method to evaluate when real-world data are suitable for causal inference.