Generalized Spatial Regression with Differential Regularization

We propose a method for the analysis of data scattered over a spatial irregularly shaped domain and having a distribution within the exponential family. This is a generalized additive model for spatially distributed data. The model is fitted by maximizing a penalized log-likelihood function with a roughness penalty term that involves a differential operator of the spatial field over the domain of interest. Efficient spatial field estimation is achieved resorting to the finite element method, which provides a basis for piecewise polynomial surfaces. The method is illustrated by an application to the study of criminality in the city of Portland, Oregon, USA.