RE: st: Modeling animal growth - nonlinear mixed model

 From "Feiveson, Alan H. (JSC-SK311)" <[email protected]> To <[email protected]> Subject RE: st: Modeling animal growth - nonlinear mixed model Date Mon, 22 Sep 2008 16:10:07 -0500

```Burt - If you want to write your own code, you could do this by
expanding the nonlinear model in a 1-step Taylor series and iterating -
for example, see Appendix 5.1 in Goldstein (below). However I agree, it
would be nice if Stata had a command to do this directly.

Goldstein, Harvey 1995. Multilevel Statistical Models, Second Ed.
London: Edward Anold Press.

Al Feiveson

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of David Airey
Sent: Monday, September 22, 2008 3:39 PM
To: [email protected]
Subject: Re: st: Modeling animal growth - nonlinear mixed model

.

I don't think there are any nonlinear mixed models in Stata (there is a
vce option for clusters in nl, but that's it). NLME in R is an option if
you have lost access to SAS, but sticking with SAS for this problem
seems best since you've done it before. Both xtmixed and gllamm are for
linear models or latent linear models (no exponentiated coefficients).
I'm sure someone will correct me if I'm wrong.

-Dave

On Sep 22, 2008, at 3:27 PM, Burt Staniar wrote:

> I have a few questions in regard to my attempts to model animal growth

> using Stata.  I am using some basic sigmoid equations to model equine
> growth.  I have previously done this in SAS, but have moved to Stata
> for most of my other work.  I would like to be able to do this work in

> Stata.  I have a longitudinal data set in a long format, with columns
> for age, day of year, weight and height.  I have successfully used the

> following model but would like to go a few steps further:
>
> nl (weight ={A=500}*(1-{b=0.9}*exp(-{k=0.003}*age))^{M=0.5}),
> cluster(animal_id)
>
> 1) How can I produce prediction or tolerance intervals for the model?
> 2) How can I produce percentile curves?
> 3) In SAS, I used nlmixed to fit both fixed and random effects.  The
> best results were placing random effects associated with animal_id on
> the A and k parameters.  Is this possible using Stata?  I have
> explored xtmixed and gllamm, but without results.
>
> Thank you for any help that may be offered.
>
> W. Burton Staniar
> Assistant Professor of Equine Nutrition The Pennsylvania State
> University
> 324 W. L. Henning Building
> University Park, PA 16802
> 814-865-0698
> [email protected]
>
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
```