A Gentle Introduction to Causal Inference: With Applications in Clinical Trials and A/B Testing

Author: Wei Liang

⚠️ Note: This blog includes works in progress and is being constantly updated.

Over a long period of time, a major focus of statistics has been on modelling associations between variables or events. However, statistical evidence based on association analysis can be misleading, because the fundamental relationship in our universe is causation rather than correlation. Association analysis refers to seeking patterns in data. For example, the occurrence of one event may often come along with another. Causal inference tries to demystify causality from data, which is not just about the patterns, but the ability to intervene on one event or variable to change the outcome of another.

The notes aim to provide a gentle introduction to causal inference and its applications in clinical trials and A/B testing, covering some elementary statistical concepts and methods such as the Neyman-Rubin potential outcomes, randomization methods, sample size determination, commonly used two-sample tests, regression analysis of experimental data. The materials are suitable for readers with a background in statistics and probability who are interested in causal inference and its connections to clinical trials and A/B testing.