Multivariate Spatial Process Models

Spatially-referenced multivariate data is becoming increasingly common. Here, we focus on data in the form of vectors observed at a finite set of spatial locations. Regression models are of interest in order to explain the response vectors as well as to predict response at unobserved locations. Such models need to capture both dependence among the components of the vectors as well as spatial dependence across the locations of the vectors. The objective of this chapter is to develop classes of models which provide the desired dependence. We consider this both formally and constructively. Constructive development is supplied through four different strategies. An example, using soil nutrient data from Costa Rica is presented.

Duke Scholars

Author Alan E. Gelfand Statistical Science