ABSTRACT
Model-based geostatistics can be used to analyze spatial data related to an underlying spatially continuous phenomenon that have been collected at a finite set of locations. Model-based geostatistics employs statistical models to capture the spatial correlation structure in the data, enabling rigorous statistical inference, and facilitating the production of spatial predictions along with uncertainty measures of the phenomenon of interest. Inference in model-based geostatistics can be performed using the integrated nested Laplace approximation (INLA) and the stochastic partial differential equation (SPDE) approaches which provide a computationally efficient alternative to Markov chain Monte Carlo methods. Briefly, this involves solving a SPDE on a discrete mesh of points and interpolating to obtain a continuous solution across the spatial domain which is calculated using INLA. Model-based geostatistics using INLA and SPDE has been employed for spatial prediction in a wide range of applications including air pollution in Italy, leptospirosis in Brazil, and malaria in Mozambique.
