Dear list,
I am looking for an implementation of random rotation matrix generation in R to
do a rotation test: I want to use the matrices to create random multivariate
normal matrices with common covariance structure and mean based on an observed
data matrix.
The rRotationMatrix-function in the mixAK-package is an option, but as far as I
can tell I need to draw rotation matrices with determinant -1 as well. Roast
and Romer in the limma-bioconductor package appear to have implemented
something similar, which seems not to be general enough for my purposes,
however. Inspired by the code in the ffmanova-rotationtest function I thought
of the following, but it appears to me that there only the covariance, not the
mean, is preserved:
#####
# a given Y has independent, multivariate normal rows
library(mvtnorm)
Y <- rmvnorm(4,mean=1:10,sigma=diag(1:10)+3)
# Generation of a set of random matrices Z
for (i in 1:10) {
# R is random matrix of independent standard-normal entries
R <- matrix(rnorm(16),ncol=4)
R <- qr.Q(qr(R, LAPACK = TRUE))
# Z shall be a random matrix with the same mean and covariance structure as Y
Z <- crossprod(R,Y)
}
#####
A suggestion for the procedure exists (in Dorum et al.
http://www.bepress.com/sagmb/vol8/iss1/art34/ , end of chapter 2.1), but a hint
to a (fast) implementation would be greatly appreciated.
Best regards and a happy new year,
Martin Krautschke
-----------------------
Martin Krautschke
Student at University of Vienna
--
Sicherer, schneller und einfacher. Die aktuellen Internet-Browser -
jetzt kostenlos herunterladen! http://portal.gmx.net/de/go/atbrowser
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
