Presenter: Edward C. Norton, PhD
Moderator: Matthew Davis, PhD
In this seminar, Dr. Edward Norton introduces entropy balancing, a statistical method for improving comparisons in observational research when randomized trials are not feasible. He contrasts it with traditional matching approaches, noting their limitations in achieving full covariate balance and maintaining sample size.
Entropy balancing reweights control observations, so their covariates exactly match the treated group, producing balanced data and more stable regression results. Through applied examples, the session demonstrates how this method reduces confounding and improves reproducibility.
Dr. Norton also highlights key considerations, including use with nonlinear models and interpretation of treatment effects, while emphasizing that entropy balancing improves—but does not establish—causal inference.
This session provides a practical, efficient approach for researchers working with real-world data.
Learn more and view other videos in the series at:
https://capra.med.umich.edu/pilot-program/curriculum-seminar-series/
#alzheimersresearch #dementiaresearch #dementia #alzheimersdisease #ADRD
