Primary Submission Category: Applications in Health and Biology
Causal dimension reduction for multiple continuous exposures with an application to environmental mixtures analysis
Authors: Thomas Hsiao, Howard Chang, Razieh Nabi,
Presenting Author: Thomas Hsiao*
Evaluating the health consequences of environmental and chemical mixtures has become a central focus in environmental epidemiology. Although substantial progress has been made in methods development, balancing flexible estimation with interpretable mixture effects remains challenging, and existing tools for drawing causal conclusions are limited. Many of these challenges stem from the difficulty of defining practical causal estimands for multidimensional, continuous exposures. We propose a sufficient dimension reduction approach that identifies low-dimensional representations preserving the causal exposure–response surface of the original mixture. We define the statistical objectives, establish theoretical properties of the resulting estimator, and evaluate its finite-sample performance through simulation. We also discuss visual and analytical strategies for interpreting the reduced dimensions. The proposed methods are compared with association-based dimension reduction techniques in simulations and illustrated through an analysis of maternal pro-inflammatory cytokine exposures and fetal inflammation during pregnancy.