Dear colleagues,
I ran into a problem with the omega matrix in nonmem 7.6 for a model I received
(simple FOCE+I advan3 trans4 nothing special).
In the model file it is specified like this (I tried different notations as
simple decimals, scientific notation etc. to no avail):
$OMEGA BLOCK(3)
0.107 ; CL
0.48E-1 0.11 ; V1
0.42E-1 0.56E-1 0.084 ; V2
However, with nonmem 7.6 I get a problem with the initial estimates of omega.
If I look at the list file I see that it is due to my omega block, which makes
sense as the correlations between these values used as initial values of etas
are not valid (outside of -1,1). However, these initial estimates are not the
ones I provided. Compare this to the same exact model ran on nonmem 7.4.4. Here
the initial omega estimates are as described in the model and it runs well.
Even in a clean folder without any files other than the dataset (csv) and the
model file (mod).
Am I doing something wrong or is this something else?
Nonmem 7.6:
INITIAL ESTIMATE OF OMEGA HAS A NONZERO BLOCK WHICH IS NUMERICALLY NOT POSITIVE
DEFINITE
0INITIAL ESTIMATE OF OMEGA:
BLOCK SET NO. BLOCK
FIXED
1
NO
0.1000E+01
0.4000E+01 0.1000E+01
0.4000E+01 0.5000E+01 0.8000E+01
Compared to nonmem 7.4.4, which seems to do this correctly:
0INITIAL ESTIMATE OF OMEGA:
BLOCK SET NO. BLOCK
FIXED
1
NO
0.1070E+00
0.4800E-01 0.1100E+00
0.4200E-01 0.5600E-01 0.8400E-01
Best regards,
Sebastiaan
Problem with initial values of the omega matrix with nonmem 7.6
Hi Sebastian,
It looks like Nonmem 7.6 is right? Notice, there is a slight difference in the 7.6 and the 7.4.4 example. There is a negative eigenvalue in the first example but not in the second.
> library(NMdata)
> om1 <- NMreadInits(lines="$OMEGA BLOCK(3)
+ 0.1000E+01
+ 0.4000E+01 0.1000E+01
+ 0.4000E+01 0.5000E+01 0.8000E+01
+ ") |>
+ dt2mat(col.value="init")
> eigen(om1)$values
[1] 13.05750679 0.02484605 -3.08235284
> om2 <- NMreadInits(lines="$OMEGA BLOCK(3)
+ 0.1070E+00
+ 0.4800E-01 0.1100E+00
+ 0.4200E-01 0.5600E-01 0.8400E-01
+ ") |>
+ dt2mat(col.value="init")
> eigen(om2)$values
[1] 0.19875329 0.06301735 0.03922935
Best,
Philip
Quoted reply history
On 2026-03-18 05:57 AM, Sebastiaan Sassen wrote:
> Dear colleagues,
>
> I ran into a problem with the omega matrix in nonmem 7.6 for a model
> I received (simple FOCE+I advan3 trans4 nothing special).
>
> In the model file it is specified like this (I tried different
> notations as simple decimals, scientific notation etc. to no avail):
>
> $OMEGA BLOCK(3)
> 0.107 ; CL
> 0.48E-1 0.11 ; V1
> 0.42E-1 0.56E-1 0.084 ; V2
>
> However, with nonmem 7.6 I get a problem with the initial estimates
> of omega. If I look at the list file I see that it is due to my omega
> block, which makes sense as the correlations between these values used
> as initial values of etas are not valid (outside of -1,1). However,
> these initial estimates are not the ones I provided. Compare this to
> the same exact model ran on nonmem 7.4.4. Here the initial omega
> estimates are as described in the model and it runs well.
>
> Even in a clean folder without any files other than the dataset (csv)
> and the model file (mod).
> Am I doing something wrong or is this something else?
>
> Nonmem 7.6:
>
> INITIAL ESTIMATE OF OMEGA HAS A NONZERO BLOCK WHICH IS NUMERICALLY
> NOT POSITIVE DEFINITE
>
> 0INITIAL ESTIMATE OF OMEGA:
> BLOCK SET NO. BLOCK
> FIXED
> 1
> NO
> 0.1000E+01
> 0.4000E+01 0.1000E+01
> 0.4000E+01 0.5000E+01 0.8000E+01
>
> Compared to nonmem 7.4.4, which seems to do this correctly:
>
> 0INITIAL ESTIMATE OF OMEGA:
> BLOCK SET NO. BLOCK
> FIXED
> 1
> NO
> 0.1070E+00
> 0.4800E-01 0.1100E+00
> 0.4200E-01 0.5600E-01 0.8400E-01
>
> Best regards,
>
> Sebastiaan
Hi
Out of curiosity, how do you submit your NONMEM jobs (eg, which command)?
Sebastien
Regeneron - Internal
Quoted reply history
________________________________
From: [email protected] <[email protected]> on behalf of
Sebastiaan Sassen <[email protected]>
Sent: Wednesday, March 18, 2026 5:57 AM
To: [email protected] <[email protected]>
Subject: [External] [NMusers] Problem with initial values of the omega matrix
with nonmem 7.6
Dear colleagues, I ran into a problem with the omega matrix in nonmem 7. 6 for
a model I received (simple FOCE+I advan3 trans4 nothing special). In the model
file it is specified like this (I tried different notations as simple decimals,
scientific
Dear colleagues,
I ran into a problem with the omega matrix in nonmem 7.6 for a model I received
(simple FOCE+I advan3 trans4 nothing special).
In the model file it is specified like this (I tried different notations as
simple decimals, scientific notation etc. to no avail):
$OMEGA BLOCK(3)
0.107 ; CL
0.48E-1 0.11 ; V1
0.42E-1 0.56E-1 0.084 ; V2
However, with nonmem 7.6 I get a problem with the initial estimates of omega.
If I look at the list file I see that it is due to my omega block, which makes
sense as the correlations between these values used as initial values of etas
are not valid (outside of -1,1). However, these initial estimates are not the
ones I provided. Compare this to the same exact model ran on nonmem 7.4.4. Here
the initial omega estimates are as described in the model and it runs well.
Even in a clean folder without any files other than the dataset (csv) and the
model file (mod).
Am I doing something wrong or is this something else?
Nonmem 7.6:
INITIAL ESTIMATE OF OMEGA HAS A NONZERO BLOCK WHICH IS NUMERICALLY NOT POSITIVE
DEFINITE
0INITIAL ESTIMATE OF OMEGA:
BLOCK SET NO. BLOCK
FIXED
1
NO
0.1000E+01
0.4000E+01 0.1000E+01
0.4000E+01 0.5000E+01 0.8000E+01
Compared to nonmem 7.4.4, which seems to do this correctly:
0INITIAL ESTIMATE OF OMEGA:
BLOCK SET NO. BLOCK
FIXED
1
NO
0.1070E+00
0.4800E-01 0.1100E+00
0.4200E-01 0.5600E-01 0.8400E-01
Best regards,
Sebastiaan
********************************************************************
This e-mail and any attachment hereto, is intended only for use by the
addressee(s) named above and may contain legally privileged and/or confidential
information. If you are not the intended recipient of this e-mail, any
dissemination, distribution or copying of this email, or any attachment hereto,
is strictly prohibited. If you receive this email in error please immediately
notify me by return electronic mail and permanently delete this email and any
attachment hereto, any copy of this e-mail and of any such attachment, and any
printout thereof. Finally, please note that only authorized representatives of
Regeneron Pharmaceuticals, Inc. have the power and authority to enter into
business dealings with any third party.
********************************************************************
Dear all,
The problem was caused by the size of the dataset, which was just over 10.000
lines (excluding censored lines). Including $SIZES LIM1=20000 fixed the problem
for me. I was put on the wrong track because it changed my omega matrix which
caused a different error, but the size was the underlying problem.
Best regards,
Sebastiaan
Quoted reply history
________________________________
From: Sébastien Bihorel <[email protected]>
Sent: Wednesday, March 18, 2026 2:52 PM
To: Sebastiaan Sassen <[email protected]>; [email protected]
<[email protected]>
Subject: Re: Problem with initial values of the omega matrix with nonmem 7.6
Waarschuwing: Deze e-mail is afkomstig van buiten de organisatie. Klik niet op
links en open geen bijlagen, tenzij u de afzender herkent en weet dat de inhoud
veilig is.
Caution: This email originated from outside of the organization. Do not click
links or open attachments unless you recognize the sender and know the content
is safe.
Did you check if the content of the .ext file at iteration 0 also reflects this
rounding issue or if this is just a .lst problem?
Regeneron - Internal
________________________________
From: Sebastiaan Sassen <[email protected]>
Sent: Wednesday, March 18, 2026 9:50 AM
To: Sébastien Bihorel <[email protected]>; [email protected]
<[email protected]>
Subject: [External] Re: Problem with initial values of the omega matrix with
nonmem 7.6
Hi Sebastien, I am use the Pirana PsN toolkit to execute the model but the same
thing occurs if I run it directly from command line and without psn. I seems
that somehow my omega block gets converted to integers in nm7. 6 for this
model. I received
Hi Sebastien,
I am use the Pirana PsN toolkit to execute the model but the same thing occurs
if I run it directly from command line and without psn. I seems that somehow my
omega block gets converted to integers in nm7.6 for this model. I received the
model from an external party and checked if I saw anything in the model file
that causes this to happen, but thus far no success. I was wondering if anyone
else ran in to this problem.
Best Sebastiaan
________________________________
From: Sébastien Bihorel <[email protected]>
Sent: Wednesday, March 18, 2026 2:42 PM
To: Sebastiaan Sassen <[email protected]>; [email protected]
<[email protected]>
Subject: Re: Problem with initial values of the omega matrix with nonmem 7.6
Waarschuwing: Deze e-mail is afkomstig van buiten de organisatie. Klik niet op
links en open geen bijlagen, tenzij u de afzender herkent en weet dat de inhoud
veilig is.
Caution: This email originated from outside of the organization. Do not click
links or open attachments unless you recognize the sender and know the content
is safe.
Hi
Out of curiosity, how do you submit your NONMEM jobs (eg, which command)?
Sebastien
Regeneron - Internal
________________________________
From: [email protected] <[email protected]> on behalf of
Sebastiaan Sassen <[email protected]>
Sent: Wednesday, March 18, 2026 5:57 AM
To: [email protected] <[email protected]>
Subject: [External] [NMusers] Problem with initial values of the omega matrix
with nonmem 7.6
Dear colleagues, I ran into a problem with the omega matrix in nonmem 7. 6 for
a model I received (simple FOCE+I advan3 trans4 nothing special). In the model
file it is specified like this (I tried different notations as simple decimals,
scientific
Dear colleagues,
I ran into a problem with the omega matrix in nonmem 7.6 for a model I received
(simple FOCE+I advan3 trans4 nothing special).
In the model file it is specified like this (I tried different notations as
simple decimals, scientific notation etc. to no avail):
$OMEGA BLOCK(3)
0.107 ; CL
0.48E-1 0.11 ; V1
0.42E-1 0.56E-1 0.084 ; V2
However, with nonmem 7.6 I get a problem with the initial estimates of omega.
If I look at the list file I see that it is due to my omega block, which makes
sense as the correlations between these values used as initial values of etas
are not valid (outside of -1,1). However, these initial estimates are not the
ones I provided. Compare this to the same exact model ran on nonmem 7.4.4. Here
the initial omega estimates are as described in the model and it runs well.
Even in a clean folder without any files other than the dataset (csv) and the
model file (mod).
Am I doing something wrong or is this something else?
Nonmem 7.6:
INITIAL ESTIMATE OF OMEGA HAS A NONZERO BLOCK WHICH IS NUMERICALLY NOT POSITIVE
DEFINITE
0INITIAL ESTIMATE OF OMEGA:
BLOCK SET NO. BLOCK
FIXED
1
NO
0.1000E+01
0.4000E+01 0.1000E+01
0.4000E+01 0.5000E+01 0.8000E+01
Compared to nonmem 7.4.4, which seems to do this correctly:
0INITIAL ESTIMATE OF OMEGA:
BLOCK SET NO. BLOCK
FIXED
1
NO
0.1070E+00
0.4800E-01 0.1100E+00
0.4200E-01 0.5600E-01 0.8400E-01
Best regards,
Sebastiaan
********************************************************************
This e-mail and any attachment hereto, is intended only for use by the
addressee(s) named above and may contain legally privileged and/or confidential
information. If you are not the intended recipient of this e-mail, any
dissemination, distribution or copying of this email, or any attachment hereto,
is strictly prohibited. If you receive this email in error please immediately
notify me by return electronic mail and permanently delete this email and any
attachment hereto, any copy of this e-mail and of any such attachment, and any
printout thereof. Finally, please note that only authorized representatives of
Regeneron Pharmaceuticals, Inc. have the power and authority to enter into
business dealings with any third party.
********************************************************************