[R-sig-ME] Question about linear mixed model

[Hung-Chih Ku]

Dear all,

I am trying to estimate variance components from a linear mixed model Y=Xa+Zb+e where Z is an nxp unstructured matrix. The following is an example of R code.


n <- 50

p <- 5

X <- runif(n)

Z <- NULL

for (i in 1:p) {

Z <- cbind(Z, rbinom(n, 2, .3))

}

a <- 0.5

b <- rnorm(p)

e<- rnorm(n)

Y <- X*a + Z*b + e

library(nlme)

group=rep(1,n)

fit <- lme(Y~X, random=list(group=pdIdent(~-1+Z)))

VarCorr(fit)

However, when n and p increase, let's say n=1000 and p=500, the lme will be taking a long time to estimate two variances (variances of b and e). Is there a way to speed up?

Thanks a lot,

Best wishes,
Hung-Chih Ku

Postdoctoral fellow
UT Southwestern Medical Center


________________________________

UT Southwestern Medical Center
The future of medicine, today.

	[[alternative HTML version deleted]]