glmmBUGS (version 2.4.2)

cholInvArray: Precision matrices to variance matrices for Winbugs output

Description

Given an array containing simulations from the posterior of a precision matrix, each individual precision matrix is converted to variances, covariances, and correlations.

Usage

cholInvArray(x, prefix = "T", chol=FALSE)

Arguments

x

An array of winbugs output, with precision matrix entries of the form "T[1,3]"

prefix

The name of the precision matrix in winbugs, the "T" in "T[1,2 ]"

chol

If TRUE, the cholesky decomposition is returned instead of the inverse

Value

An array with the third dimension's precision matrix entries changed to

"sdT[i,i]"

for the standard deviation of component i

"covT[i,j]"

for the covariance between i and j

"corrT[i,j]"

for the correlations between i and j

Details

Inverts the matrices with the cholesky decomposition, but operating on all matrices simultaneously using array arithmetic.

Examples

# NOT RUN {
# create a random positive definite matrix by 
# generating a lower triangle
  N=4
  lmat = diag(runif(N, 1, 10))
  thetri = lower.tri(lmat)
  lmat[thetri] = rnorm(sum(thetri), 0, 2)
#  precmat = solve(lmat %*% t(lmat))
  precmat = solve(lmat %*% t(lmat))
 
# put this matrix into an array 
  precarray = array(c(precmat), dim=c(1,1,length(precmat)))
  dimnames(precarray) = list(NULL, NULL, 
    paste("T[", rep(1:N, N), ",", rep(1:N, rep(N,N)), "]",sep="") )

# invert it with cholInvArray and the solve function
  cholInvArray(precarray)[1,1,]
  # the off diagonals of solve(precmat) should be 
  # the covT elements of cholInvArray(precarray)
  solve(precmat)
  # the standard deviations in cholInvArray(precarray) should be the 
  # root of the diagonals of solve(precmat)
  sqrt(diag(solve(precmat)))
# }