From: "Yungwei Hsu" 
Subject: [NMusers] Model Validation
Date: Tue, 19 Mar 2002 13:24:20 -0500

Dear all,

I have a 3-compartment nonlinear model based on only 7 subjects' data.
Could somebody tell me the best way to do model validation? I had tried
Jackknife but failed because the standard error became too big when I took
out any subject's data to run NONMEM. And I don't think bootstrapping will
work.

By the way, what is the minimal subjects number to do bootstrapping?

Thanks,

Yung-Wei Hsu MD
Research Fellow
Duke University Medical Center


*******

From: "Joern Loetsch" 
Subject: Re: [NMusers] Model Validation
Date: Tue, 19 Mar 2002 21:04:05 +0100
          1

What about getting independent data from somewhere and look how good your
model predicts them?
Jorn

*******

From:  "Piotrovskij, Vladimir [PRDBE]" 
Subject: RE: [NMusers] Model Validation
Date: Wed, 20 Mar 2002 09:16:16 +0100


I wonder how many random effects do you have in the model? I don't think you can estimate
more than one or two having the data from 7 subjects only.

Best regards, 
Vladimir


*******