From: "Eyas Abu-Raddad" <RADDADE@rx.uga.edu>
Subject: Nested random effects with NONMEM?
Date: Sat, 11 Mar 2000 22:27:51 -0500
Can NONMEM handle nested random effects? That is nonlinear models in which random effects occur in a multilevel structure. For example, PK data from each of several rats within each of several batches or litters.
Eyas AbuRaddad
Department of Pharmaceutical and Biomedical Sciences
University of Georgia
Nested random effects with NONMEM?
From: LSheiner <lewis@c255.ucsf.edu>
Subject: Re: Nested random effects with NONMEM?
Date: Mon, 13 Mar 2000 10:43:04 -0800
At present, it is oriented to just 2 levels - within individual noise and between individual variation. With some difficulty it can he "fooled" into modeling a third intermediate level, usually inter-occasion variation. In your case, you would want litter-to-litter variation, rat-to-rat, and within-rat. It could be done if the max. number of rats within litter (n1) were small and the number of parameters in the within-rat model (n2) were also small, as the way one fools NONMEM into creating the middle layer is to use a different random effect for each parameter on each "occasion", or in your case, rat-within-litter. This uses a total of n2(n1+1) random effects. This total is limited to no greater than 30 in most current installations of NONMEM (note that because of the nested structure, this implies far fewer distinct random effect parameters, however).
LBS.
--
_/ _/ _/_/ _/_/_/ _/_/_/ Lewis B Sheiner, MD (lewis@c255.ucsf.edu)
_/ _/ _/ _/_ _/_/ Professor: Lab. Med., Bioph. Sci., Med.
_/ _/ _/ _/ _/ Box 0626, UCSF, SF, CA, 94143-0626
_/_/ _/_/ _/_/_/ _/ 415-476-1965 (v), 415-476-2796 (fax)
From: "Pascal Girard" <PGIRARD@pharsight.com>
Subject: Re: Nested random effects with NONMEM?
Date: Mon, 13 Mar 2000 22:58:37 +0100
Some time ago, Vladimir Piotrovski. indicated that nlme Splus function new release (also available in freeware R) could handle nested random effects. Because of the limitations of NONMEM indicated by Lewis, Silvy Laporte and myself tested nlme via different simulations using 3 and 4 random nested models.
We found that 3 (and even more) nested levels was easy to implement, except that you need to write your PK model. Results was very close from the one we got with NONMEM using the implementation suggested by Lewis. For models with 4 nested random effects that NONMEM could not handle we found, using simulations, unbiased and precise results with nlme. In addition nlme appeared very fast even with this complex variance structure.
Pascal Girard
pgirard@pharsight.com
From: "Sarah Marston" <smarston@statsci.com>
Subject: FW: Nested random effects with NONMEM?
Date: Sun, 19 Mar 2000 15:37:21 -0500
For those interested in using the S-Plus nlme, a book 'Mixed Effects Models in S and S-Plus', by Pinheiro, Jose et al will be available from Springer Verlag (ISBN: 0387989579) by the end of this month. You can find more info and order from amazon.com.
Sincerely,
Sarah Marston