Hands-on Tutorials
Causal Inference in Data Science: A/B Testing and the need for Marginal Structural Modeling
Required Adjustment of A/B Tests via G-Methods in the Presence of Informative Censoring
1: Background and Motivation
Causal Inference is of interest to a wide range of practitioners including Statisticians, Data Scientists, Machine Learning Scientists, and other Computational Researchers. Recovery of unbiased estimates of Causal Effects is at times a tough task. In my previous pieces on Doubly Robust Estimation and G-Estimation of Structural Nested Models, we discussed leveraging G-Methods in non-randomized settings. We also discussed issues with M-Bias and confounding identification in non-randomized settings.
Contrary to popular belief, recovery of valid estimates of Causal Effects can be equally as difficult (or sometimes even unidentifiable and impossible) with A/B Testing (aka randomized trials). Regarding the use of A/B Testing in industry settings, I’ve noticed an often over-simplified view of randomized trial designs; a view of A/B Testing that corresponds to “idealized randomized trials”. For A/B Tests conducted at scale, particularly those that are longitudinal and carried-out over lengthy periods…
