Abstract: Directed Acyclic Graphs (DAG) are visual representations facilitating the rigorous investigation of causal relationships. DAGs have gained traction across multiple disciplines in recent years as useful tools for social science researchers for determining whether causal effects may be identified in observational data and to infer the implications of causal models. This presentation begins by providing an overview of the counterfactual model of causal inference and its application to observational studies. We then proceed to introduce DAGs and explain how they are useful in examining relationships between variables (causation, confounding, and endogenous selection). Finally, we explore how DAGs can be used to derive associations between two variables depending on the conditioning of intermediary variables. This presentation will serve to introduce DAGs to a non-specialist audience.
David Monaghan is a PhD candidate in Sociology at the Graduate Center, CUNY. His research is mostly in the areas of higher education and social stratification. He and Paul Attewell are authors of Data Mining: A Gentle Introduction, under contract at UC Press.
Dirk Witteveen is a PhD student in Sociology at the Graduate Center, CUNY. He holds a BA in Sociology from the University of Amsterdam and an MA in Sociology of Education from New York University. His research interests include the transition of higher education to the labor market, the economic returns to educational credentials, and the position of children of immigrants in the labor market. He is currently working on the application of Hidden Markov Models to predict student’s completion perspectives in 4-year colleges with Paul Attewell.
Darren Kwong is a PhD student in Sociology at the Graduate Center, CUNY. He holds a BA in Political Science from Vassar College. He is the Project Coordinator of the CUNY Data Mining Initiative. His research interests include race and education, youth movements, and emotions. He is currently working on a data mining paper examining interactions in data with Paul Attewell.