Hi everybody,
I need help in writing a statistical function for bootstrap. Suppose m is a
matrix with n cols and p rows, my original data. What I want to do is a
bootstrap (using boot from package boot) on eigenvectors from a PCA done on m
with a statistic function calculating the eigenvector bootstrap error ratio.
If R = number of bootstrap replicates, then my function should look something
like this where m.i is the ith bootstrap sample matrix (of course this is wrong,
but I wrote it to give an idea what I actually want)
pcasig <- function(m) {
for (i in 1:R) {
pca.i <- prcomp(m.i)
eigen.i <- pca.i$rotation
estim <- sum(eigen.1+….+eigen.R)/R
se <- ((sum((eigen.i-estim)^2))/(R-1))^0.5
t.i <- estim/se }
}
Thanks for your consideration,
Monica
_________________________________________________________________
07
[[alternative HTML version deleted]]
