[R] nlme question

[Doran, Harold]

I think the authors are mistaken. Sigma is random error, and due to its
randomness it cannot be systematically related to anything. It is this
ind. assumption that allows for the likelihood to be expressed as
described in Pinhiero and Bates p.62. 

If you are finding that sigma is systematically related to a fixed
effect, then your model is misspecified and there are omitted
characteristics that need to be accounted for. 

-----Original Message-----
From: Wassell, James T., Ph.D. [mailto:jtw2 at cdc.gov] 
Sent: Thursday, November 17, 2005 11:52 AM
To: Doran, Harold; r-help at stat.math.ethz.ch
Subject: RE: [R] nlme question

Thank you for taking the time to think about my problem. 

The reference states:  "The covariance structure must be considered,
because for unbalanced data the estimates" (i.e. mu, sigma and tau hats)
"are not typically independent." Page 105.  It would be nice to simply
assume zero covariance terms, but the authors reject this
simplification.