From: peter.bonate@quintiles.com

Subject: Lag Time Problem

Date: Wed, 4 Apr 2001 12:04:19 -0500

 

I am having a problem with ADVAN4 TRANS4. I am running a 2-compartment model with proportional error model. Up to this point, everything has been running smooth. When I add a lagtime (ALAG1) to the model, I get the following error message:

 

MONITORING OF SEARCH:

 

0PROGRAM TERMINATED BY OBJ

ERROR IN CELS WITH INDIVIDUAL 1 ID= .10000000E+01

WEIGHTED SUM OF "SQUARED" INDIVIDUAL RESIDUALS IS INFINITE

MESSAGE ISSUED FROM ESTIMATION STEP

AT 0TH ITERATION, UPON EVALUATION OF GRADIENT

 

So what I did next was to add an additive error term to make an additive and proportional error model. The model converges with a decrease in the objective function of 13 points, but with R-matrix singularity - the additive EPS(2) term is 1E-13.

 

So next I tried fixing the EPS(2) term to a very small value, 0.0001. I still get R-matrix singularity.

 

Can anyone suggest something here.

 

Thanks.

 

Peter Bonate



*****

From: "Gibiansky, Ekaterina" <gibianskye@globomax.com>

Subject: RE: Lag Time Problem

Date: Wed, 4 Apr 2001 13:24:35 -0400

 

Peter,

 

did you try to make an initial value for LAG smaller than the time of first observed non-zero concentration? It may help at 0 iteration, though this error can re-appear later. If it does, ETA for lag may help, though it is a shaky thing

 

Katya

________________________

Ekaterina Gibiansky, PhD

Senior Scientist

 

GloboMax LLC

7250 Parkway Drive, Suite 430

Hanover, MD 21076

Voice (410) 782-2234

FAX (410) 712-0737

E-mail: gibianskye@globomax.com