Weather2vec: Representation Learning for Causal Inference with Non-Local Confounding in Air Pollution and Climate Studies

Mauricio Tec, James G. Scott, Corwin Zigler

Keywords: spatial confounding, deep learning, climate change, causal inference, potential outcomes, interference

TL;DR: This paper defines non-local confounding, which affects causal inference with spatial data, and investigates a U-net methodology to address it in air pollution studies.

Abstract: Non-local confounding (NLC) can bias the estimates of causal effects when treatments and outcomes of a given unit are dictated in part by the covariates of other units. This paper first formalizes the problem of NLC using the potential outcomes framework, providing a comparison with the related phenomenon of causal interference. Then it investigates the use of neural networks -- specifically U-nets -- to address it. The method, termed "weather2vec", uses balancing scores to encode NLC information into a scalar or vector defined for each observational unit, which is subsequently used to adjust for NLC. We implement and evaluate the approach in two studies of causal effects of air pollution exposure.