Data Fusion, Simulations, and Bias Analysis for Credible Causal Inference

The Department of Epidemiology Seminar Series

Open to the public.

Speaker:

Onyebuchi A. Arah, MD, PhD
Professor, Department of Epidemiology
Affiliated Professor, Department of Statistics
Associate Dean, Graduate Division
UCLA Fielding School of Public Health

Abstract: The growing interest in causal inference has rightly led to increased work in at least three areas that may seem disparate at first glance. First, there is the recent interest in data fusion spurred by transportability. Second, quantitative bias analysis is gaining renewed attention as its tools become more accessible for use in empirical studies. Third, simulations are increasingly used for estimation, estimator performance evaluation, incorporating uncertainty, and teaching in causal inference. Using illustrative examples of uncontrolled confounding in epidemiologic studies, this talk will lay out important connections between data fusion, simulations, and bias analysis, and how they can be used to make causal inference more credible. In modern causal modeling, data fusion is critical for designing studies, combining multiple data sources, evaluating counterfactuals, transporting effects, and conducting bias analysis. Simulations aid fuller implementations of data fusion and bias analysis. Finally, bias analysis, which can be cast as an exercise in data fusion, can use simulations to examine the sensitivity of study results to violations of causal assumptions or identifiability conditions. Indeed, it will become clear that fusion, especially when combined with simulations, is a general framework for causal inference tasks such as study design, combining data sources, estimating counterfactuals, effect estimation, conducting what-if virtual experiments, heterogeneity assessment, transportability, and bias analysis.

Bio: Onyebuchi (Onyi) Arah, MD, MSc, MPH, DSc, PhD is a Professor of Epidemiology at Fielding School of Public Health, an affiliated Professor at the Department of Statistics, and an Associate Dean of Graduate Education at the Graduate Division, all at University of California, Los Angeles (UCLA). He is the President-Elect of the Society for Epidemiologic Research (SER). His research interests include epidemiologic methodology, causal inference, pediatric and perinatal epidemiology, computational epidemiology, health services research, and social epidemiology. He has received several research, teaching and service awards including the Council of Science Editors Award for outstanding contributions to global health policy and practice, the European Society for Philosophy, Medicine and Health Care’s First Prize for Young Scholars under age 35 who have made innovative contributions to the philosophy of medicine and health care, the Causality in Statistics Education Award from the American Statistical Association, an Honorary Skou Professorship from Aarhus University in Denmark, the Academic Council Chairs Award for Mid-Career Leadership from the University of California Systemwide Academic Senate, and the Outstanding Contributions to Epidemiology Award from the American College of Epidemiology.