Title:
A model for large multivariate spatial datasets
Abstract:
Multivariate spatial modeling is a rapidly growing field, but most extant models are infeasible for use with massive spatial processes. In this work we introduced a highly flexible, interpretable and scalable multiresolution approach to multivariate spatial modeling. Relying on compactly supported basis functions and Gaussian Markov random field specifications for coefficients results in efficient and scalable calculation routines for likelihood evaluations and co-kriging. We analytically show that special parameterizations approximate popular existing models. Moreover, the multiresolution approach allows for arbitrary specification of scale dependence between processes. We illustrate our approach through Monte Carlo studies to illustrate implied stochastic behavior and test our ability to recover scale dependence, and moreover examine a complex large bivariate observational minimum and maximum temperature dataset over the western United States.
Biography
Soutir Bandyopadhyay earned a doctorate in statistics at Texas A&M University, a master’s degree in statistics at the Indian Statistical Institute in New Delhi and a bachelor’s degree in statistics at St. Xavier’s College in Calcutta. He has also been a visiting scientist at the Computational and Information Systems Laboratory at the National Center for Atmospheric Research studying climate models. Bandyopadhyay’s area of expertise is time series and spatial statistics. He has published his work in the Journal of Time Series Analysis, Journal of Computational and Graphical Statistics, the Journal of the Royal Statistical Society, Series B and the Annals of Statistics.
