This is the program for the 2010 Joint Statistical Meetings in Vancouver, British Columbia.

Abstract Details

Activity Number: 179
Type: Contributed
Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
Sponsor: Section on Bayesian Statistical Science
Abstract - #308191
Title: Adaptive Gaussian Predictive Process Model for Large Spatial Data Sets
Author(s): Rajarshi Guhaniyogi*+ and Andrew Finley and Sudipto Banerjee and Alan E. Gelfand
Companies: University of Minnesota and Michigan State University and University of Minnesota and Duke University
Address: 410,6TH STREET SE, APT NO.314, MINNEAPOLIS, MN, MN55414, United States
Keywords: Bayesian inference ; Gaussian process ; Hierarchical modelling ; Markov chain Monte Carlo ; Predictive Process ; Spatial data analysis
Abstract:

Over time,hierarchical models using MCMC have become increasingly popular for spatial modelling due to more flexibility and power compared to classical methods.Fitting these models involve matrix decompositions which might be infeasible for large spatial data.Banerjee et al(2008) use Predictive process model which relies on the choice of set of knots. In this paper we devise two different methods for knot selection,one employing reversible jump MCMC technique which allows the selection of optimal set of knots,and the second using preferential sampling method(Diggle,2009)to construct a joint prior distribution on the location of knots and random effects,thereby selecting the optimal set of knots after running MCMC.Both methods considerably decrease the number of knots needed to carry out proper inference, thereby reducing the computational complexity of the Predictive process model.


The address information is for the authors that have a + after their name.
Authors who are presenting talks have a * after their name.

Back to the full JSM 2010 program




2010 JSM Online Program Home

For information, contact jsm@amstat.org or phone (888) 231-3473.

If you have questions about the Continuing Education program, please contact the Education Department.