Introduction to Causal Graphs (Directed Acyclic Graphs, DAGs)

Session Submission Type: Short Course Full Day

Session Description

Causal diagrams, or directed acyclic graphs (DAGs), are among the most widely used tools in causal inference. DAGs offer an accessible and intuitive way to explain data-generating processes and research designs. They help clarify theory, communicate assumptions, and formulate estimation strategies. A variety of free software tools now allow users to easily draw DAGs, assess whether quantities of interest are causally identified, and obtain causal bounds that are robust to common complications such as selection or mismeasurement.

In this full-day short course, we provide a comprehensive introduction to DAGs and their use in applied political science. The course begins by covering basic principles of causal graphs. It then illustrates how canonical research designs and common challenges in applied work can be concisely represented in the language of causal graphs. Substantial time is devoted to interactive practice for translating plain-language theory into caual diagrams and formal assumptions. The course concludes with a survey of freely available software tools for practitioners. Details are provided in the outline below.

Introduction to DAGs
- Counterfactuals
- Covariate adjustment
- Colliders
- D-separation

DAGs for canonical research designs and common challenges
- Canonical designs: SOO, IV, DID, RDD
- Common challenges: mismeasurement, missingness, selection
- Game theoretic models

Interactive exercises
- Translating theory into DAGs
- Practice and feedback on drawing DAGs

Software tools
- Graphical interfaces for drawing DAGs
- Assessing causal identification
- Bounds for partially identified quantities

Sub Unit

• Pre-Conference Short Courses

Chair

• Dean Knox University of Pennsylvania

Presenters

• Guilherme Jardim Duarte Princeton University

• Jonathan Mummolo Princeton University

• Cyrus Samii New York University

• Tara L Slough New York University