The goal of ghInf is to provide tools for inference for general Gaussian hierarchical models. Note that since this package is for personal use, all observations are assumed to be 0 for now, and the mean of the root parameter also assumed to be zero (subject to updating in the future). The functions can be found in the R folder. This includes functions which:
Generate the precision matrix for centered 2- or 3-level Gaussian hierarchical models:
- centered_precgen2.R
- centered_precgen3.R
Generate the precision matrix for any 2- or 3-level Gaussian hierarchical models:
- precgen2.R
- precgen3.R
Calculate the fill-in ratio of the calculated Cholesky factor relative to the optimal fill-in for both spam and Matrix packages, and for different label permutations:
- centered_fillin2.R
- centered_fillin3.R
- fillin2.R
- fillin3.R
A Gibbs sampler for centered 2- or 3-level Gaussian hierarchical models:
- centered_gibbs2.R
- centered_gibbs3.R
A helper function for permuting the label of the precision matrix into a depth-first order from a lexographical order:
- depthfirst.R
You can install the released version of ghInf by using the following code.
library(devtools)
install_github("kwajiehao/ghInf")You may also use the githubinstall package.
library(githubinstall)
githubinstall("ghInf")Test examples which show how the functions work can be found in the "tests" folder.