Doubly-robust covariate balancing weights for causal inference and transportability
Kevin Josey · 2020
Nearly every discussion of a clinical research study includes an evaluation of internal and external validity. While randomized clinical trials can overcome several threats to internal validity, they may be prone to poor external validity. Conversely, large prospective observational studies sampled from a broadly generalizable population may be externally valid, yet susceptible to threats to internal validity, particularly confounding. Thus, methods that address confounding and enhance transportability of study results across populations are essential for internally and externally valid causal inference, respectively. In this dissertation, we develop a class of weighting estimators that solve both of these problems in separate contexts. We also propose a solution to the scenario where both internal and external validity must be addressed concomitantly. The weighting methods we present are doubly-robust, which allows for consistent estimation even when some components of the statistical model are misspecified. Several simulation experiments are conducted to compare our proposed solutions with alternative methods found in the literature. We also apply our results to real-world data examples to help illustrate these concepts and their importance in medical and scientific research.