On the goodness-of-fit of generalized linear geostatistical models

TL;DR

This paper introduces a generalized coefficient of determination for generalized linear geostatistical models, enhancing interpretability and variable contribution assessment in spatial prediction, with applications demonstrated in river-blindness mapping.

Contribution

It generalizes Zhang's coefficient of determination for geostatistical models, providing a more intuitive measure for model fit and variable importance assessment.

Findings

The generalized coefficient of determination improves interpretability.

Application to river-blindness mapping demonstrates practical utility.

Method applicable to any generalized linear mixed model.

Abstract

We propose a generalization of Zhang's coefficient of determination to generalized linear geostatistical models and illustrate its application to river-blindness mapping. The generalized coefficient of determination has a more intuitive interpretation than other measures of predictive performance and allows to assess the individual contribution of each explanatory variable and the random effects to spatial prediction. The developed methodology is also more widely applicable to any generalized linear mixed model.