About the Package: {geoCount}
This package provides a variety of functions to analyze and model
geostatistical count data with generalized linear spatial models,
including,
1) simulate and visualize the data;
2) posterior sampling with robust MCMC algorithms (in serial or
parallel way);
3) perform prediction for unsampled locations;
4) conduct Bayesian model checking procedure to evaluate the
goodness of fitting;
5) conduct transformed residual checking procedure.
In the package, seamlessly embedded C++ programs and parallel computing
techniques are implemented to speed up the computing processes.
Available for both Linux/Unix and Windows systems.
Package Dependency
• R (>= 2.12.0)
• C++ compiler: for example GNU Compiler Collection (GCC)
• GNU Scientific Library (GSL): a numerical library for C and C++
• LAPACK and BLAS (or ATLAS): libraries for numerical linear algebra operations
• Standard development tools such as make etc.
• R packages
– Rcpp (>= 0.9.4): provides a C++ API as an extension to the R system
– RcppArmadillo (>= 0.2.19): integration for “Armadillo” which is a templated C++ linear algebra library
– {coda}: for Markov chain diagnostics
– {distrEx}: for calculating Hellinger and Kolmogorov distances between two distributions
– {reldist}: for calculating the relative density
– {multicore, snow, snowfall}: for parallel computing
Caution: {Rcpp} and {RcppArmadillo} are required to install
{geoCount}. Other R packages are only required when you use
the corresponding functions in the package.
Documentation
Manual: [pdf]
Tutorial: "How to Use {geoCount}" [pdf]
|
Code
[R]
Download
Source code: [geoCount_1.130409.tar.gz]
Binary for Windows: for 32-bit R-3.0.0 [geoCount_1.130409.zip]
*See the tutorial for how to install the package.
Methodology
• Data and Model Specification
• Robust MCMC Algorithms
• Bayesian Model Checking
• Transformed Residual Checking
• Techniques for building the package: API between R and C++; parallel computing in R
*See my dissertation for details.