Design of experiments (DOE) and randomized control trials are the gold standard for determining causal relationships, and JMP is the gold standard for DOE software. Unfortunately, many studies with causal inference objectives cannot be run as DOEs due to an inability to randomly assign treatments because of practical or ethical constraints. The requirement is to remove the impact of confounding variables that contribute to bias in the observational data classes.

We provide an overview of current promising causal inference methods to include propensity scores, matching algorithms, and other statistically based approaches to include draft FDA guidance to industry for real world data/evidence. We demonstrate several procedures in JMP using a retrospective study database that supports multiple clinical research efforts for the Cardiothoracic Surgery Program at Houston VA Medical Center. We finish by providing recommendations, tips, and pitfalls for the practitioner to consider when causal association is the goal and DOE is not an option.