Dr. Andrew Yarger, Purdue University - Asymmetries in multivariate spatial and spatio-temporal covariances

Abstract: Asymmetries in dependence may exist in both multivariate spatial and spatio-temporal environmental data. Flexible covariance models are needed to capture these features, and most existing approaches are based on shift or delay-based asymmetries. We first introduce an approach to incorporate asymmetry in multivariate spatial models through the spectral representation of the covariance. Examples are discussed for multivariate Matérn, squared-exponential, and Cauchy covariance models. We then show how these models can be used straightforwardly to introduce asymmetries in space-time covariance functions.