Improving Efficiency in Estimating Transported Average Treatment Effects

Kara Rudolph, PhD

Associate Professor of Epidemiology - Columbia University

WHEN: Wednesday, September 30, 2026, from 3:30 to 4:30 p.m.
WHERE: Hybrid | 2001 McGill College Avenue, Room 1140; Zoom
NOTE: Kara Rudolph will be presenting in person

Abstract

The effect of a treatment or exposure on an outcome may differ from one sample or population to another. Such treatment effect heterogeneity drives concerns about external validity—specifically, in the presence of such treatment effect heterogeneity, a treatment effect estimated in a study sample may not apply to a target population. There can be multiple drivers of such heterogeneity, which we review. We focus on two scenarios here. First, the simplest scenario where the only source of heterogeneity is the differential distribution of effect modifiers across populations, and second, where the sources of heterogeneity also include differences in the distribution of mediating variables. Even semiparametric efficient estimators of transported average effects can have large variability and wide confidence intervals in scenarios where there is limited overlap between the distribution of these variables across populations. This motivates the need to develop transport estimators with improved precision. In this paper, we show that the precision of transported estimates may be improved by considering two types of covariate dimension reduction; we develop an estimator for each approach. First, we propose a so-called “collaborative” estimator that reduces the dimension of the transport problem to improve efficiency. Second, we construct two other semiparametric estimators for cases where the causal structure underlying effect heterogeneity is known.

Speaker Bio

I am an epidemiologist with research interests in developing and applying causal inference methods to better understand how to prevent and treat substance use disorders and prevent violence. My current work focuses on understanding issues related to data support and complex exposures and developing and applying methods for transportability and mediation to understand mechanisms relevant for drug use disorder prevention and treatment in various target populations. More generally, my work on generalizing/ transporting findings from study samples to target populations and identifying subpopulations most likely to benefit from interventions contributes to efforts to optimally target available policy and program resources. I completed a PhD in Epidemiology and an MHS in Biostatistics from the Johns Hopkins Bloomberg School of Public Health and was a Robert Wood Johnson Foundation Health and Society Scholar.