Essential Concepts for Causal Inference in Randomized Experiments and Observational Studies
Speaker: Donald Rubin, John L. Loeb Professor Statistics, Harvard University
Abstract: There are several essential concepts for understanding causal inference and for implementing it in practice. These concepts were clearly formulated only recently, beginning in the early twentieth century, and are critically important to keep in mind when trying to understand the causal effects of past interventions or new proposed interventions, especially when the effects of such interventions are subtle because they are relatively small or their size depends on characteristics of the objects to which they are applied. Some historical connections and the importance of modern computing will be emphasized. Also, the reasons for the inapposite focus on “regression” based methods, which still dominates some practice, will be discussed.