Leveraging Observational Outcomes to Improve the Generalization of Experimental Results

Melody Huang (University of California, Los Angeles), Erin Hartman (University of California, Los Angeles), Naoki Egami (Columbia University) and Luke Miratrix (Harvard University)

*Award for Best Graduate Student Poster - Methods*

Abstract: Randomized control trials are often considered the gold standard in causal inference due to their high internal validity. However, generalizing experimental results to a target population can be a challenge in social and biomedical sciences. Recent papers clarify assumptions necessary for generalization and develop various weighting estimators for the population average treatment effect (PATE). Unfortunately, in practice, many of these methods result in large variance and little statistical power, thereby limiting the value of the PATE inference. In this study, we show that when information about the outcome variable is available in observational population data, we can leverage this information to improve the efficiency of many existing popular methods without making additional assumptions. We empirically demonstrate the efficiency gains through simulations and apply our proposed method to the Get Out the Vote data.

View Poster in a New Tab


37136c44caf9656ed77bda51cc40a97a