Functional inverted Wishart for Bayesian multivariate spatial modeling with application to regional climatology model data.

ABSTRACT: Modern environmental and climatological studies produce multiple outcomes at high spatial resolutions. Multivariate spatial modeling is an established means to quantify cross-correlation among outcomes. However, existing models typically suffer from poor computational efficiency and lack the flexibility to simultaneously estimate auto- and cross-covariance structures. In this article, we undertake a novel construction of covariance by utilizing spectral convolution and by imposing an inverted Wishart prior on the cross-correlation structure. The cross-correlation structure with this functional inverted Wishart prior flexibly accommodates not only positive but also weak or negative associations among outcomes while preserving spatial resolution. Furthermore, the proposed model is computatio

SUBMITTER: Duan LL 

PROVIDER: S-EPMC5863270 | biostudies-literature | 2017 Nov

REPOSITORIES: biostudies-literature