An application to Galicia pollution data and a model for preferential sampling

Abstract

In geostatistics, it is commonly assumed that sample locations are equally spread over the observed region. Furthermore, it is assumed that the point process for data locations does not depend on the data process. One has preferential sampling when the last assumption fails. We address this problem by proposing a model-based approach, for stationary Gaussian spatial processes. This new parametric model is founded on a flexible class of log-Gaussian Cox processes.

In this talk, we investigate an application example to pollution data, where we believe there is a clear rationale for the sampling being preferential.