Conference Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 362 - Causal Inference for Undergraduates: Teaching Correlation Does Not Imply Students Understand Causal Inference
Type: Invited
Date/Time: Wednesday, August 10, 2022 : 8:30 AM to 10:20 AM
Sponsor: Section on Statistics and Data Science Education
Abstract #320489
Title: Covariate Balance as Paramount to Causal Inference
Author(s): Kari Lock Morgan*
Companies: Penn State University
Keywords: causal inference; causality; statistics education; randomization; propensity scores

Whether teaching introductory students or PhD students, whether discussing randomized experiments or observational studies, the core fundamental idea I want to convey to students about causal inference remains the same: for causal inference, we want to compare groups that are comparable to begin with. In other words, we want all baseline variables (covariates) to be similarly distributed (balanced) between the groups being compared. In experiments, we achieve this through randomization and elements of experimental design. In observational studies, we can balance observed covariates by comparing matched units that are similar, looking within similar subclasses, or weighting units to have similar weighted covariate distributions; all of these can be illustrated with a single covariate or extended with propensity scores. While the depth and topic coverage will vary from course to course, the goal of balancing covariates, and helping students understand why this covariate balance is so crucial, remains the same.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2022 program