9I制作厂免费

Nima Hejazi, PhD

Assistant Professor, Department of Biostatistics
Harvard Chan School of Public Health

WHEN: Wednesday, October 22, 2025, from 3:30 to 4:30 p.m.
WHERE: Hybrid | 2001 9I制作厂免费 College Avenue, Rm 1140;
NOTE:听Nima Hejazi will be presenting virtually from Boston
听

Abstract

In contemporary studies in environmental health and epidemiology, observational data feature structure that poses several distinct challenges for causal inference. Distinct problems include the presence of continuous exposures; many measured putative confounders of the causal relationship of the exposure and outcome; and network interference, wherein the exposure of one unit may affect the outcome of its neighbors. Each problem requires care to address. We introduce a novel intervention scheme鈥攖he induced modified treatment policy鈥攄esigned to aid in the identification of the causal effect of intervening on a continuous exposure while simultaneously addressing network interference between study units; this addresses the first and third of the challenges noted above. Building on a fast-growing literature in causal machine learning, we formulate asymptotically (semi-parametric) efficient estimators of the causal effects of induced modified treatment policies, which allow for the use of non-parametric regression and/or machine learning techniques for estimation of nuisance functions; this addresses the second challenge noted above by allowing for the data analyst to flexibly adjust for many putative confounders while acknowledging the limitations of domain knowledge in practice. In numerical experiments, we illustrate how induced modified treatment policies eliminate causal (i.e., identification) bias due to network interference. We apply the proposed methodology to evaluate the causal effect of zero-emission vehicle uptake on air pollution in California, strengthening evidence from prior analytic studies.

Speaker Bio

I am an assistant professor of biostatistics at the Harvard Chan School of Public Health. My research interests primarily center on causal inference, semi-parametric statistics (particularly causal or de-biased machine learning), assumption-lean and non-parametric inference, and computational statistics. My methodological work is usually motivated by applied science investigations in the infectious disease sciences, the study of chronic diseases, and cancer science. I鈥檓 also deeply interested in both open-source software and high-performance computing for the statistical sciences鈥攖o push the boundaries of statistical methodology and to promote transparency and reproducibility in the practice of applied statistics and statistical data science.听

Website: