Causal Effects Under Spatial Confounding and Interference (Short Paper)

Abstract

Spatial causal inference is an emerging field of research with wide ranging areas of applications. As a key methodological challenge, spatial confounding and spatial interference can compromise the performance of standard statistical inference methods. In the current literature, there is a lack of appreciation of the connections between spatial confounding and interference. This could potentially lead to overspecialized silos of research. Therefore, we need further research to bridge such gaps theoretically, and to find creative solutions for complex spatial causal inference problems. This short paper offers a brief demonstration: It discusses the connections between spatial confounding and interference. An illustrative simulation study shows how commonly used approaches compare across four test scenarios. The simulation study is discussed with an emphasis on the promising performance of counterfactual prediction based inference methods.