PROC NLMIXED

[ramvieira]

Dear all,

It is past the point that a Continuous AutoRegressive Correlation (CAR) should have been accounted for in PROC NLMIXED. In Vonesh Book, SAS programming sentences for auto regressive correlation for qualitative factors are introduced, but non is for CAR. It would be great if SAS teams account for it!

[sbxkoenk]

Hello,

Are you talking about this book?

Generalized Linear and Nonlinear Models for Correlated Data: Theory and Applications Using SAS
by Edward F. Vonesh (Author)
2012

Why do you think you cannot work with a continuous-time autocorrelation correlation / error structure in PROC NLMIXED?
You can! (It's not because it's not in the book that it cannot be done)

Koen

[ramvieira]

Dear One

Yes, I am talking about the book.

Please, let me know a reference about a continuous auto-regressive
correlation in NLMIXED. It will help me a lot!

Kind regards and thank you very much,

Ricardo Vieira

[sbxkoenk]

Hello,

I am quite sure I had found a paper yesterday using CAR within PROC NLMIXED.
I am searching for 10 à 15 minutes now (to find it again), but cannot spot it anymore.

Unfortunately I do not have an example of this type of model using PROC NLMIXED readily available. 

Maybe @jiltao can help you out?

By the way, is this on pharmaco-kinetics and compartmental models?

If yes, this is FYI. ... ... Currently, NLMIXED is indeed the best option for nonlinear mixed effect models; however, it’s support for ordinary differential equations used in pharmacokinetic models is limited to 1, 2, and 3 compartment models with specific dosing (aka forcing functions). 

I believe SAS is working on adding support for more general pharmacokinetic models.

Thanks,

Koen