The Use of Propensity Scores to Estimate Sample Selection Error in Observational Data
Eloise Kaizar, Ohio State University 
*Taylor R. Pressler, Ohio State University College of Medicine 

Keywords: Propensity scores, randomized controlled trials, bias

While randomized controlled trials (RCT) are considered the “gold standard” for clinical studies, the use of exclusion criteria may impact the external validity of the results. It is unknown whether estimators of effect size are biased by excluding a portion of the population. However, it may be possible to use data from observational studies to estimate a difference between the population average treatment effect (PATE) of the included and excluded portions of the population, the sample selection error (SSE). We propose an estimator for the SSE and use simulation to study its properties while considering a non-constant treatment effect. We find that a doubly robust estimator that uses both propensity scores and a model for the outcome generally outperforms an estimator that solely relies on the use of propensity scores, even when model elements are misspecified.