Re: RE: error message 0VAR-COV OF SOME OBS ESTIMATED TO BE 0 - a bug?
Hello Robert,
Thank you for your help. I figured out the potential reason. I took care of
an additive component of the error. It seems like it is wrong to assume that
M2-M4 take care of the additive component. It is wrong to assume that the
additive component is trivial when M2-M4 are used. Also, the same SD should be
used in both the standard and the M2-M4 portions of the objective function.
It will be nice if there is a sorce, which explaines the error messages. The
error messages can be puzzeling.
Kind regards,
Pavel
Quoted reply history
----- Original Message -----
From: "Bauer, Robert"
Date: Thursday, February 10, 2011 7:59 pm
Subject: RE: [NMusers] error message 0VAR-COV OF SOME OBS ESTIMATED TO BE 0 - a
bug?
To: [email protected], Martin Johnson
Cc: [email protected]
> Pavel:
> If you do not mind sharing your control stream file, and if necessary,
> your data file (or a truncated version of it), perhaps you could
> send it
> to me. I may be able to determine if there is a way to modify the
> control stream code to avoid these errors, or if there is a bug.
>
> Robert J. Bauer, Ph.D.
> Vice President, Pharmacometrics
> ICON Development Solutions
>
> Tel: (215) 616-6428
> Mob: (925) 286-0769
> Email: [email protected]
> Web: www.icondevsolutions.com
>
>
>
>
>
>
>
> ________________________________
>
> From: [email protected] [mailto:owner-
> [email protected]]On Behalf Of [email protected]
> Sent: Thursday, February 03, 2011 9:32 AM
> To: Martin Johnson
> Cc: [email protected]
> Subject: Re: [NMusers] error message 0VAR-COV OF SOME OBS
> ESTIMATED TO
> BE 0 - a bug?
>
>
> Thank you Martin,
>
> It was helpful and educational, but did not resolve the issue.
> I just
> used a very different code and datasets. The only common part
> was
>
> $EST MAXEVALS=9999 SIG=5 NOABORT PRINT=1 SORT
> METHOD=IMP CTYPE=3 NITER=1000 ISAMPLE=300 POSTHOC INTERACTION
>
> There are no warnings or errors. Newertheless, the code creates
> 2 out
> of 4 requested output files. Both output files are truncated, i.e.
> after a certain subject and a timepoint there are no output
> records. It
> does not depend on the number of added dummy time points
> (MDV=1). What
> can it be?
>
> Thanks,
> Pavel
>
> ----- Original Message -----
> From: Martin Johnson
> Date: Friday, January 28, 2011 1:52 am
> Subject: Re: [NMusers] error message 0VAR-COV OF SOME OBS
> ESTIMATED TO
> BE 0 - a bug?
> To: [email protected]
>
> > Dear Pavel,
> > Just adding my few cents..
> > I understand that you are very experienced NMUSER, so this
> mail
> > may not
> > add anything new to you.
> >
> >
> > Error message on FNLETA comes from the posthoc step- "An error
> > message
> > from FNLETA indicates that NONMEM has experienced difficulties
> > obtaining posthoc etas"
> > Error message on CELS "usually" says that there are some
> "zero"
> > predictions.
> > So, in this way..error message makes some sense...
> >
> > However, i have not used IMP with M3 and Laplacian, so i can
> not
> > comment
> > on that.
> >
> > One thing may be...model also tries to predict the pk or pd
> for
> > zero
> > time points...that could cause this problem....if so, CALLFL=0
> > in the
> > $ERROR may be of some use
> >
> > Regards,
> > Martin Johnson,
> > PhD Student,
> > University of Groningen,
> > The Netherlands
> >
> >
> >
> > On 01/28/2011 12:31 AM, [email protected] wrote:
> > > Hello Nonmem Users,
> > >
> > > I am getting an error message:
> > > 0PROGRAM TERMINATED BY FNLETA
> > > ERROR IN CELS WITH INDIVIDUAL 2 ID= 1.10050000000000E+01
> > > 0VAR-COV OF SOME OBS ESTIMATED TO BE 0
> > > MESSAGE ISSUED FROM TABLE STEP
> > >
> > > The estimate step is as follows:
> > > $EST MAXEVALS=9999 SIG=5 NOABORT PRINT=1 SORT
> > > METHOD=IMP CTYPE=3 NITER=1000 ISAMPLE=300 POSTHOC INTERACTION
> > > MSFO=KI4H004D.MSF LAPLACIAN
> > > The code also implements M3. The code runs fine, but it does
> > not
> > > create the output data files specified in the table
> > statements. I
> > > tried to remove some table statements, changed initial
> > esimates, added
> > > the additive component to the error model. Nothing helps.
> > When I do
> > > not use M3 and Laplacian, I am getting the output files. I
> > used M3and
> > > Laplacian with some other drugs, but I did not use it in
> > combination
> > > with METHOD=IMP. Can it be a NONMEM bug? The eror does not
> > make sense.
> > > Thanks,
> > > Pavel
> >
> >
>
>