A new way of analyzing malaria data: A non-stationary geostatistical modeling approach

Abstract Geostatistical models are widely used to analyze malaria data, and obtain spatial predictions at un-sampled locations based on the First Law of Geography (close things in space are more similar than distant things). When environmental covariates affect not only the mean of the underlying process under investigation but also its covariance structure, stationary models for spatial prediction are questionable. In this paper, we illustrate how to incorporate spatially referenced environmental risk-factors into the covariance function to model non-stationary patterns of malaria risk. Specifically, we demonstrate the suggested modelling framework with a case study of malaria prevalence in Mozambique where we compare a non-stationary model with dependence structure governed by precipitation to the standard stationary model. Results reveal that non-stationary geostatistical modeling approaches are more useful to model the non-stationary patterns of malaria. Further the predication malaria risk map based on the non-stationary modeling approach show that in many part of the country (especially in the eastern part) malaria prevalence is above 30%. The demonstrated non-stationary modeling approach will play a great role for malaria elimination.