VEST Workshop with Dr. Peter Steiner

Graphical Models for Causal Inference (this event is only open to VEST students)

  • -
  • Ridley 302

The workshop is intended as an introduction to the theory and application of graphical models (also referred to as causal graphs or directed acyclic graphs—DAGs). The major goal is to learn about the foundations of graphical models and to use them as powerful tools for designing, analyzing, and critically assessing evaluation studies. The first part of the course starts with a thorough introduction to graph terminology, causal graphs and structural causal models (SCMs), with a focus on drawing graphs from subject matter theory, understanding collider variables and collider bias, and using graphical concepts like d-separation to derive testable implications of the presumed causal model. Then, graphical models are used to precisely define causal effects and to discuss identification criteria like the backdoor or frontdoor criterion. The identification results will be linked to nonparametric and parametric estimation strategies of causal effects. The workshop also highlights useful applications of causal graphs including omitted variables bias and covariate selection issues, alternative causal identification strategies (gain scores/difference-in-differences, instrumental variables), the design of causal studies, counterfactuals and mediation analysis. The computing software R will be used for graphical and statistical analyses of real and simulated data to demonstrate the entire research process from translating subject matter theory into causal graphs to the analysis of data. For the graphical analyses, the R packages dagitty, ggdag, and causaleffect will be introduced.

Speaker Bio
A headshot of Peter Steiner, VEST guest speaker

Peter M. Steiner is a Professor in the Department of Human Development and Quantitative Methodology at the University of Maryland, College Park. His primary research interest is in causal inference, covering the design and analysis of experiments and quasi-experiments, causal replication designs, graphical models, and factorial surveys. His research has appeared in such journals as Psychological Methods, Multivariate Behavioral Research, Journal of Educational and Behavioral Statistics, Evaluation Review, Sociological Methods & Research, Journal of Causal Inference, or the Journal of the American Statistical Association. He has taught introductory and advanced graduate courses on graphical models for causal inference at the University of Wisconsin-Madison and the University of Maryland. He has also held workshops on graphical models and quasi-experimental designs in the United States, Europe and Asia. In 2019, he received the Causality in Statistics Education Award of the American Statistical Association. Peter M. Steiner holds a PhD in Statistics from the University of Vienna, Austria.

Event Information

Event Sponsor

  • Virginia Education Science Training (VEST) Fellowship Program