Bayesian Spanning Tree-based Multivariate Spatial Model

Aim to develop an interpretable and computational feasible method for highly multivariate (large p) and huge (large n) spatial data by utilizing relationships among variables in an inter-variable graph.

Consider a minimum spanning tree as the backbone of the inter-variable graph.

Transform a multivariate process into multiple bivariate processes by exploiting variable-level conditional independence specified by the minimum spanning tree.