Inferring causality is central to many quantitative studies in social science. A large number of analytical methods have been developed to infer causal dependence from observational data, including propensity score matching, instrumental variable designs, interrupted time-series designs, and many others. Unfortunately, the assumptions and limitations of these methods can be difficult to explain and reason about.
This 2-day (9-hour) tutorial introduces participants to causal graphical models, a powerful formalism developed within computer science and statistics that simultaneously provides: 1) a unifying formal framework for understanding and explaining specific methods for causal inference; 2) a practical tool for representing and reasoning about the implications of particular causal models; and 3) powerful algorithmic methods for learning complex causal models from data and reasoning about their implications. This tutorial assumes only a basic understanding of probability and statistics and no knowledge of programming. Participants familiar with experimental and quasi-experimental designs will gain a new understanding of the benefits and assumptions of these methods, and participants without that knowledge will learn about multiple methods for inferring causality within a single unifying framework.
