Computational Methods for Large Spatial Temporal Data Sets

• Activity Number: 6
• Type: Invited
• Date/Time: Sunday, August 4, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Huiyan Sang*+ and Bohai Zhang and Jianhua Z. Huang
• Companies: TAMU and Texas A&M University and Texas A&M University
• Keywords: Covariance function ; Gaussian Process ; Geostatistics ; Reducend rank ; Sparse matrix ; MCMC

Abstract

There has been much interest in recent years in the statistical modeling of data collected over space and time. With large spatial temporal data sets, the implementation of spatial temporal modelling typically poses significant computational challenges. We will discuss several possibilities of extending the full-scale covariance approximation approach to the spatial temporal modeling of large data sets. We consider both the discrete-indexed and the continuous-indexed temporal settings. In particular, we present several modifications of the full-scale approximation method aiming to make good use of the specific space-time model structure. The methods are illustrated with simulation studies and climate model outputs.