TITLE: Introduction to Instrumental Variables
INSTRUCTOR: Nandita Mitra, PhD, UPenn and Jason Roy, PhD, Rutgers University
MODERATOR: Kalyan Ghosh



The goal of this tutorial is to provide a practical introduction to instrumental variables and how they are used for the analysis of observational studies.  We will begin with a brief overview of causal inference concepts including simple directed acyclic graphs, the potential outcomes framework, average causal effects, and identifying assumptions. We will then briefly review standard approaches to analyzing observational data such as propensity score matching and inverse probability of treatment weighting. We will then motivate the need for using instrumental variable (IV) approaches to account for unmeasured confounding. We will define what an IV is and describe the underlying assumptions. We will provide examples of common IVs used in the clinical literature and discuss their limitations. We will also demonstrate how to use IVs to estimate average treatment effects using two-stage models, such as two-stage predictor substitution and two-stage residual inclusion, and how to interpret the results. Strengths and limitations of these methods will be discussed as well. Throughout, we will use examples from our own work in cancer comparative effectiveness studies. We will demonstrate how to implement IV estimation using the R package ivreg. The instructors have a long-standing interest in developing causal inference methodology and are co-directors of the Center for Causal Inference (with Dr. Dylan Small):

Instructors’ Biography:

Jason Roy, PhD is Professor and Chair of the Department of Biostatistics and Epidemiology at Rutgers University, Co-Director of the biostatistics core for the New Jersey Alliance For Clinical and Translational Science, and Co-Director of the Center for Causal Inference. He has expertise in Bayesian methods, causal inference, and missing data. His primary recent methodological research has focused on developing flexible Bayesian models for complex observational data, especially from large healthcare databases. He has collaborated in a wide variety of clinical research areas, including chronic kidney disease, chronic viral hepatitis infection, and HIV.



Nandita Mitra, PhD is Professor of Biostatistics, Vice-Chair of Faculty Professional Development, Chair of the Graduate Group in Epidemiology and Biostatistics, and Co-Director of the Center for Causal Inference at the University of Pennsylvania. Her primary methodological research focuses on propensity score and instrumental variables approaches to the analysis of observational data and causal inference approaches to cost-effectiveness estimation. Her collaborative research areas include cancer outcomes, cancer genetics, health policy, and health economics.

This entry was posted in . Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *