[R-sig-ME] nonlinear mixed longitudinal model

[Ken Beath]

R, unlike SAS, doesn't provide functions for fitting random effects  
models with arbitrary likelihoods. The only solution is to write your  
own, using one of the integral approximations: Laplace, gauss-Hermite  
or adaptive Gauss-Hermite. Which one is appropriate depends on several  
factors. If speed doesn't matter, and clusters are not too large,  
Gauss-Hermite can be programmed very rapidly. A possibility is to  
modify the lme4 code but that doesn't look easy.

Ken

On 27/04/2009, at 1:12 PM, Lee, MinJae wrote:

> Hello,
>
>
> I was trying to convert the Proc nlmixed SAS to nlme in R.
>
> especially, I'm considering the approach of Thiébaut and  
> *Gadda( *Mixed models for longitudinal left-censored repeated  
> measures in SAS program:
>
> proc nlmixed
>
> parms alpha=1 beta1=1 beta2=1   s11=0.5 s22=0.5 se=1;
> bounds 0<=s11, 0<s22, 0<se;
> eta=alpha+beta1*&gp+beta2*day+ u1+day*u2;
> ll=(1-&cens)*(log((1/(sqrt(2*(3.14159)*se)))*exp((-1/ 
> (2*se))*((&coags-eta)**2))))
> +(&cens)*(log(probnorm((&coags-eta)/(sqrt(se)))));
> model &coags ~ general(ll);
>
> random u1 u2 ~ normal([0,0],[s11,0,s22]) subject=subjectid;
>
>
>
> I know there are a lot of differences between them(nlmixed vs nlme),  
> but I really need to convert this program to R.
>
> could you please give any advise for me?
>
> is there any appropriate packages in R for this model?
>
>
>
> Thank you so much in advance,
>
> Minjae Lee
>
> _______________________________________________
> R-sig-mixed-models at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>