|
Abstract:
|
Basis functions have long been an integral part in specifying multivariate models, and have natural connections to spatial processes through matrix decompositions, Green's functions, and spectral representations. Process convolutions provide yet another recipe for producing flexible spatial models, particularly useful for non-standard models, such as multivariate, non-Gaussian, and non-stationalry processes. This talk will look at historical uses of process convolutions, connections to other basis approaches, and uses outside of the spatial modeling domain.
|
