- This event has passed.
Biostatistics Seminar: Abhirup Datta, PhD Candidate, University of Minnesota
February 15, 2016 @ 3:30 pm - 5:00 pm
University of Minnesota
“Nearest Neighbor Gaussian Process Models for Massive Spatial
and Spatiotemporal Data”
Abstract: Gaussian process (GP) models are widely used for analyzing space and space-time indexed data from forestry, environmental health, climate sciences etc. However, traditional GP models entail computations that become prohibitive for modern geostatistical datasets with large number of spatial or temporal locations. In this talk, I will present our proposed Nearest-neighbor Gaussian process (NNGP) models which provide a highly scalable alternative for fully model based inference for massive spatial and spatio-temporal datasets. NNGP is a well-defined spatial process and can be used as a sparsity-inducing prior for spatial or spatio-temporal random effects within a rich hierarchical modeling framework. Matrix-free Markov chain Monte Carlo (MCMC) algorithms for NNGP deliver massive scalability. NNGP effectively reproduces the corresponding inference from traditional (but highly expensive) GP models. I will also discuss applications of NNGP to massive scale prediction of forest biomass and analysis of air pollution data. Light refreshments will be served at 3:00 pm in room 1629 SPH I.