[R-sig-ME] user defined covariance structure in lme4 /nlme (Tapan Mehta)

[Tapan Mehta]

Thanks Stephan. I appreciate your response. I agree with you about
convergence issue. I ran into this when trying on a slightly more complicated
pedigree data and had to use compound
symmetry matrix.

 I have tried using corSymm but it does not give me correct random
effects estimates as compared to what I get through SAS. I am trying to
estimate the genetic and error variance and based on a simulated data
(through SAS) with genetic variance 7 and error variance 2 whereas the
results I obtain from the code below gives 7.58 and 1.38. 

migGdata<-c(0.0,0.5,0.5)
cs1Symm<-corSymm(migGdata,form=~1 | FAM_ID)
cs2Symm<-Initialize(cs1Symm,data=migTrio)
test<-lme(y ~ x1 + x2, random=~1|FAM_ID,correlation=cs2Symm,data=migTrio)

Let G = k*C, where G is the variance of the random effects, C is a
symmetric positive definite matrix, and k a parameter. We would like to
supply lme /lmer  with C (user defined covariance structure) and let it
estimate k. I am also trying to use the pdSymm method to implement this
but I am running into this error:

 > mt
     [,1] [,2] [,3]
[1,]  1.0  0.0  0.5
[2,]  0.0  1.0  0.5
[3,]  0.5  0.5  1.0
pd2 <- pdSymm(mt,form=~1|FAM_ID,name=c("1","2","3"))
testRes<-lme(Y~X1+X2,random=pd2,data=migTrio)

Error in getGroups.data.frame(dataMix, groups) : 
Invalid formula for groups

Please let me know if any of you can suggest what I need to fix to solve this problem.

Thanks,

Tapan


----- Original Message ----
From: Stephan Moratti <moratti at med.ucm.es>
To: r-sig-mixed-models at r-project.org
Sent: Wednesday, April 16, 2008 3:05:20 AM
Subject: [R-sig-ME] user defined covariance structure in lme4 /nlme (Tapan Mehta)

Hi Tapan,

The following line using lme from the nlme package should work:

testRes <- lme(Y ~X1+X2,random=~1|FAM_ID,correlation=corSymm(form = ~1|FAM_ID),data=fam_dat)

testRes <- lme(Y ~X1+X2,random=~1|FAM_ID,correlation=corSymm(),data=fam_dat) should be equal.

I hope this helps. Sometimes I had the problem that the REML did not converge using the corSymm(), whereas using a compound symmetry structure worked.

Best,

Stephan




-- 
*Stephan Moratti, PhD/
/**/see also: http://web.mac.com/smoratti/
/*Centro de Tecnología Biomédica CBT,
Universidad Politécnica de Madrid,

en la actualidad (currently at) en el
Centro de Magnetoencefalografía Dr. Perez Modrego,
Universidad Complutense de Madrid,
Faculdad de Medicina,
Pabellón 8,
Avda. Complutense, s/n,
28040 Madrid,
Spain,

email: moratti at gbt.tfo.upm.es <mailto:moratti at gbt.tfo.upm.es>
        moratti at med.ucm.es
Tel.:    +34 91 394 2292
Fax.:   +34 91 394 2294
*/
/*

    [[alternative HTML version deleted]]