fillin3: centered_fillin3

In kwajiehao/ghInf: Inference for Gaussian hierarchical models

Description

This function returns the ratio of the number of zeros obtained from the cholesky decomposition of the specified precision matrix over the optimal cholesky fill-in for a general 3-level gaussian hierarchical model. If the cholesky decomposition is optimal, the number returned is 1, and if it is sub-optimal the number returned is greater than 1. The function also returns the time taken for the cholesky decomposition.

Usage

fillin3(method, permute_method = "none", i, J, K, permute = TRUE, ...)

Arguments

method	determines which linear algebra package to use
permute_method	determines how the precision matrix is permuted before use with the chosen linear algebra package
i	number of nodes at level 1
J	vector specifying the number of children nodes in level 2 per node at level 1
K	vector specifying the number of children nodes in level 3 per node at level 2
permute	determines whether or not to use default row permutation algorithm before matrix factorization to reduce fill-in
...	for other params used in centered_precgen3

Details

A choice of 2 packages is provided: spam and Matrix. A choice of 3 permute methods is provided: no permutation, random permutation, and depth-first permutation.

Value

the fill-in ratio with respect to the optimal fill-in, and the time taken for the cholesky decomposition

Examples

i <- 2
J <- c(1,2)
K <- c(1,2,3)
fillin3(method = 'matrix', permute_method = 'none', i = i, J=J, K=K)

kwajiehao/ghInf documentation built on May 7, 2019, 10:58 a.m.