[R-sig-ME] user defined covariance structure in lme4 /nlme

[Tapan Mehta]

Hi all,

I am trying to use the lme4 /nlme  package to fit a mixed model on pedigree data.I would really appreciate if anyone of you could forward an example of inputting a user defined covariance structure for the random effects. I tried using the 'start' argument in the lmer method of the lme4 package but was not successful. I am not sure either whether the 'start' argument will help me achieve this.

 I have also tried using the nlme package to achieve this. Initialize.corSymm was the method I used and implemented it in the way described below

lData<-c(0.0,0.5,0.5)
cs1Symm<-corSymm(ldata,form=~ fam_dat$FAM_ID)
cs2Symm<-Initialize(cs1Symm,data=fam_dat)
testRes <- lme(Y ~ X1 + X2, random=~ 1|FAM_ID,correlation=cs2Symm,data=fam_dat)

I did not find any example in the documentation of the package that shows this. It would really be helpful if somebody could point me out any existing documentation on how to fit a user defined covariance structure either using lme4 or the nlme package.

Please let me know if you need any further information.

Thanks,

Tapan