Abstract

There are many competing approaches to modeling spatial phenomenon such as Gaussian Processes, Spatial Nearest Neighbor Gaussian Processes, Stochastic Partial Differential Equations, Integrated Nested Laplace Approximations and Tensor Product approaches. While these all have their advantages and disadvantages there does not seem to be a definitive winner. As part of a TIES working group we examined these methods in a large scale simulation study. This talk we look at a few of the simulated datasets and compare the results of each of the methods. In addition, we compare and contrast these methods against a harmonic analysis (sometimes called Fourier) approach.