Spatial+: a novel approach to spatial confounding

Emiko Dupont,Simon N. Wood,Nicole Augustin

DOI: https://doi.org/10.1111/biom.13656

2020-09-20

Abstract:In spatial regression models, collinearity between covariates and spatial effects can lead to significant bias in effect estimates. This problem, known as spatial confounding, is encountered modelling forestry data to assess the effect of temperature on tree health. Reliable inference is difficult as results depend on whether or not spatial effects are included in the model. The mechanism behind spatial confounding is poorly understood and methods for dealing with it are limited. We propose a novel approach, spatial+, in which collinearity is reduced by replacing the covariates in the spatial model by their residuals after spatial dependence has been regressed away. Using a thin plate spline model formulation, we recognise spatial confounding as a smoothing-induced bias identified by Rice (1986), and through asymptotic analysis of the effect estimates, we show that spatial+ avoids the bias problems of the spatial model. This is also demonstrated in a simulation study. Spatial+ is straight-forward to implement using existing software and, as the response variable is the same as that of the spatial model, standard model selection criteria can be used for comparisons. A major advantage of the method is also that it extends to models with non-Gaussian response distributions. Finally, while our results are derived in a thin plate spline setting, the spatial+ methodology transfers easily to other spatial model formulations.

Methodology,Statistics Theory,Applications

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the significant bias problem caused by the collinearity between covariates and spatial effects in the spatial regression model, namely the so - called "spatial confounding". Specifically, when adding spatial effects to the model, the impact estimates of covariates may be severely distorted, so that the results depend on whether spatial effects are included. This phenomenon is particularly evident in forestry data modeling for evaluating the impact of temperature on tree health. Reliable inferences become difficult because the results are affected by whether spatial effects are incorporated. Currently, there is insufficient understanding of the mechanisms behind spatial confounding, and the treatment methods are limited. Therefore, this paper proposes a new method - "spatial+" - aiming to avoid this bias problem by reducing the collinearity between covariates and spatial effects.