From: Qun Jiao <Qun.Jiao@dupontpharma.com>

Subject: Rounding Error 134

Date: 13 Aug 2001 15:18:17 -0400

 

Hi,

 

I am the new Nonmem user. I have a problem that need your help. See the attached data set "DRUGB". The first time, I used control file "DRUGB-1.txt". A complete output ("DRUGB-1.out") comes out, even the objective function and estimated parameters did not look so good. The second time, I use the predicted parameters from DRUGB-1 (see control file DRUGB-2.txt), the run did not complete with a rounding error. DRUGB-2.out.txt Also, when I tried to use exact predicted para. number from DRUGB-1, it won't let me run through until I limit the decimal places for Theta, Omega and Sigma. I did change several parameters, but always get the same "rounding error, error 134" message. Did anybody ever encounter the same problem? Does anybody know what "error 134" means?

By the way, I used "Digital Visual Fortran Optimizing Compiler Version 6.5, Copyright 2000".

 

Thanks in advance for your help.

 

QJ

 

*****

 

 

Date: Tue, 14 Aug 2001 16:56:36 +1200

From: Nick Holford <n.holford@auckland.ac.nz>

Subject: Re: Rounding Error 134

 

Qun,

 

The rounding error message is very common. I expect to get it most of the time when developing NONMEM models. It is especially likely when you use the final estimates from a run that minimized successfully. NONMEM needs to 'feel its way around' a bit to get sig digs. If you try starting with estimates displace by say 10% from the final values you may have more luck.

 

I think your model is perhaps too constrained. You only have ETA on CL, V2 and ALAG1. You should try putting ETA on the other PK parameters (especially KA) and also use a BLOCK structure to capture the covariance.

 

You have multiple doses for each subject so you might also want to try estimating between occasion variability.

 

Finally, graphs of the fits dont show much of a 2 cpt behaviour. You might find that a one compartment model is adequate.

 

Nick

--

Nick Holford, Divn Pharmacology & Clinical Pharmacology

University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand

email:n.holford@auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556

http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm

 

*******

 

From: TRANCHAND Brigitte <tranchan@lyon.fnclcc.fr>

Subject: Re: Rounding Error 134

Date: Tue, 14 Aug 2001 02:23:50 -0700 (PDT)

 

The rounding error 134 message is very common, and can be solved by changing

initial estimates, or OMEGA, SIGMA.

 

Your OMEGA and SIGMA are too constrained.

 

I tried this values (change for initial LAG) and it runs successful.

$THETA (0,39,200) (0,325,10000) (0,8,200) (0,105,1000) (0,2,2000) (0,1,5)

$OMEGA 5 5 5

$SIGMA 0.5 3

 

After a look to your data, I think that it is a 2-comp decline.

Moreover, why do you use FOCE method ? Results are fine with FO.

 

Hope this help

 

Dr Brigitte Tranchand

Lab Pharmacocinétique

Centre Léon-Bérard

69373 Lyon Cedex 8

France

 

Tel : 33 4 78 78 27 59

Fax : 33 4 78 78 27 58

e-mail : tranchan@lyon.fnclcc.fr