| 
|
Readme file
SERIES
B
 Statistical
Methodology
Gaussian predictive process models for large spatial data sets, by S. Banerjee, A. E. Gelfand, A. O. Finley and H. Sang, pages 825–848
Description
Following Section 2.2 in the paper, this program fits the univariate predictive process model with a stationary and isotropic Exponential spatial correlation function. The program is written in C++ and leverages R's standalone math library (Rmath) and Intel's Math Kernel Library. Please see the Makefile for and example of how to link the associated libraries. In addition to these external libraries the main function defined in pp_geostat.cpp requires supporting functions in dataMatrix.cpp and kvpar.cpp. Program parameters are defined in pfile. Synthetic data can be generated using the R program simulate.R Once the program is compiled and run using the synthetic data and associated pfile, the result is a file of MCMC samples with rows corresponding to the model's regression coefficients $\bm{beta}, partial sill $sigma^2$, nugget $\tau^2$, and spatial range $\phi$. The R program coda.R can be used to summarize these MCMC samples.
Files:
coda.R -- summarize MCMC samples
dataMtrx.cpp -- needed for pp_geostat.cpp
dataMtrx.h -- needed for pp_geostat.cpp
kvpar.cpp -- needed for pp_geostat.cpp
kvpar.h -- needed for pp_geostat.cpp
Makefile -- example Makefile for pp_geostat.cpp
pfile -- parameter file read by pp_geostat.cpp
pp_geostat.cpp -- main function of fitting predictive process geostatical model
simulate.R -- produces simulated data for pp_geostat
Sudipto Banerjee
Division of Biostatistics
School of Public Health
University of Minnesota
Mayo Mail Code 303
Minneapolis
MN 55455-0392
USA
E-mail: sudipto@biostat.umn.edu
|
Journals
SERIES
A
Statistics
in Society
SERIES
B
Statistical
Methodology
SERIES C
Applied Statistics
SERIES D
The
Statistician
|
