Simplifying Matching Methods for Causal Inference (University of Minho)

This talk explains how to make matching methods for causal inference easier to use and more powerful. Applied researchers commonly use matching methods as a data preprocessing step for reducing model dependence and bias, after which they use whatever statistical procedure they would have without matching, such as regression. They routinely ignore the requirement that all matches be exact, and also commonly use ad hoc analyses that iterate between formal matching methods and informal balance and sample size checks. The talk describes 3 papers which (1) offer the first comprehensive theory of statistical inference that does not require changing these widely used and understood procedures (http://j.mp/MatchTheory); (2) demonstrate that the most commonly used method, propensity score matching, should not be used, as it increases imbalance, model dependence, and bias and is dominated by most other methods (http://j.mp/PScore); and (3) develop a method for optimizing balance and matched sample size simultaneously by quickly estimating the optimal frontier without ad hoc iterations, thus increasing speed and ease of use (http://j.mp/MatchFrontier). (Based on joint work with Stefano Iacus, Chris Lucas, Rich Nielsen, and Giuseppe Porro.)