B0474
Title: Adjusting for nonlocal spatial confounding with U-nets in studies of meteorology and air pollution
Authors: Corwin Zigler - University of Texas at Austin (United States) [presenting]
Mauricio Tec - University of Texas at Austin (United States)
Abstract: Causal effects of spatially-varying exposures on spatially-varying outcomes can be subject to nonlocal confounding, which exists when treatments and outcomes for an index unit are dictated in part by covariates of other (perhaps nearby) units. We offer a deep-learning approach to encode nonlocal covariate information into a vector defined for each observational unit that can be used to adjust for nonlocal confounding. The approach is based on a type of convolutional neural network, called a U-net, that leverages the idea that regional confounding information can be processed in a manner similar to the information contained in an image. We illustrate the approach in two studies of causal effects of air pollution exposure, where meteorology is an inherently regional construct that threatens causal estimates with both local and nonlocal (regional) information. We illustrate the ability of the proposed U-net representation to capture relevant nonlocal confounding information that cannot be fully characterized with simple functions of local and regional meteorological covariates.
