Causal Inference for Multilevel Data - Online Course

ABSTRACT

Over the past several decades, a paradigm shift in statistics has transformed causal inference methods in social science and medicine. The key idea is simple: each study participant possesses a potential outcome under each possible intervention. Causal effects are then comparisons of these potential outcomes.

This way of defining causal effects has generated a host of new concepts and methods in randomized experiments, natural experiments, and observational studies. Our aim in this workshop is to apply these concepts and methods in multilevel settings. In such settings, causal effects are generated by the actions of heterogeneous agents (e.g., teachers, social workers, police, and physicians) operating in varied organizational settings (e.g., schools, clinics, precincts, hospitals). The fundamental challenge, but also a key focus of interest, is the heterogeneity of causal effects that arise in multilevel settings.

• The first session of the seminar will focus on randomized trials and how we might define and estimate the average and variance of “intent-to-treat” effects, “complier average causal effects,” and mediation effects, with a focus on multi-site trials and trials in which whole clusters (e.g., classrooms, schools, or hospitals) are the unit of random assignment. We’ll challenge conventional thinking on fixed versus random effects models and propose alternatives.
• The second session will extend these methods for multilevel observational studies using analysis of covariance, matching, and weighting.
• The third session will consider natural experiments in multilevel settings using regression discontinuity (RDD) and difference-in-difference (DID) methods.

In each session, we’ll consider modeling decisions, software choices, and assumption checks. Illustrative data sets include the National Head Start Impact Study, the Tennessee Class Size experiment, the Early Childhood Longitudinal Study, and High School Math reform in Chicago, with references to other similar studies.

Prerequisites: Knowledge of multiple regression. A basic introduction to methods of causal inference is helpful but not required.

Relevant Reference

Raudenbush, S.W., Schwartz, D., (2020). Randomized experiments in education, with implications for multilevel causal inference. Annual Review of Statistics and Its Application. 7:1, 177-208.

This Distinguished Speaker Series seminar will consist of three hours of lecture and Q&A, held live* via the free video-conferencing software Zoom.

*The video recording of the seminar will be made available to registrants within 24 hours and will be accessible for four weeks thereafter. That means that you can watch all of the class content and discussion even if you cannot participate synchronously.

Closed captioning is available for all live and recorded sessions. Captions can be translated to a variety of languages including Spanish, Korean, and Italian. For more information, click here.