An Improved Doubly Robust Estimator Using Partially Recovered Unmeasured Spatial Confounder
Studies in environmental and epidemiological sciences are often spatially varying and observational in nature with the aim of establishing cause and effect relationships. One of the major challenges with such studies is the presence of unmeasured spatial confounders. 'spatial confounding' is the phenomenon in which the spatial residuals are correlated to the spatial covariates in the model, when unaccounted for it can lead to biased causal effect estimates. This paper develops a novel method that adjusts for the spatial confounding bias under a spatial-causal inference framework when the treatment is binary. By combining tools from spatial statistics and causal inference literature, we propose a method that reduces the bias due to spatial confounding. We partially recover the unmeasured spatial confounder using the spatial residuals and propose an improved doubly robust estimator based on it. Through simulation studies, we demonstrate that the proposed doubly robust estimator outperforms the existing methods and has the lowest bias and close to nominal coverage in most scenarios. Finally, we implement our method to estimate the effect of installing SCR/SNCR NOx emission control technologies on ambient ozone concentrations.
READ FULL TEXT