NONMEM Users Network Archive

Hosted by Cognigen

[NMusers] RE: nonmem question regarding condition number

From: Herbert STRUEMPER <herbert.x.struemper_at_gsk.com>
Date: Thu, 24 Mar 2016 13:22:40 +0000

Dear Laureen,

I recommend that you determine with exploratory plotting how strong the evi=
dence for the 3-compartment behavior is (eg. in semi-log plots: is 3 cpt b=
ehavior present in most subjects? are inflection points reasonably collocat=
ed? how do slope changes relate to measurement noise? etc.).

If you do see strong evidence for 3 cpt behavior and have a biological rati=
onale for a third compartment, I wouldn't just drop the 3 cpt model becaus=
e of a high condition number in the model you tested, especially since you =
said that precision did not deteriorate.
The fact that the 2 cpt model was borderline ill-conditioned (998 ~ 1000) w=
ith serially sampled IV data suggests that the 2 cpt model was already over=
-parameterized. My recommendation is to simplify the 2 cpt model (eg. by s=
implifying random structure, looking for large entries in correlation matri=
x or RSEs, ...) and then run the comparison again.

                Herbert



From: owner-nmusers_at_globomaxnm.com [mailto:owner-nmusers_at_globomaxnm.com] On=
 Behalf Of Denney, William S.
Sent: Thursday, March 24, 2016 6:53 AM
To: L.A. ten Berg - Lammers; 'nmusers_at_globomaxnm.com'
Subject: [NMusers] RE: nonmem question regarding condition number


EXTERNAL
Hi Laureen,

Large condition numbers (typically interpreted as >1000) indicate that two =
or more parameters in the model are highly correlated in their covariance a=
nd that the model parameters are difficult to identify. Given your descrip=
tion below, I would not suggest using the 3-compartment model (even with it=
s 180 point reduction in OFV) because two of your parameters are so highly =
correlated that the condition number is reaching the level of computational=
 precision (2E16 is very large for a condition number). I'd also look at t=
he NONMEM output to see if the 3-compartment model had any errors around th=
e convergence.

Thanks,

Bill

From: owner-nmusers_at_globomaxnm.com<mailto:owner-nmusers_at_globomaxnm.com> [ma=
ilto:owner-nmusers_at_globomaxnm.com] On Behalf Of L.A. ten Berg - Lammers
Sent: Thursday, March 24, 2016 6:10 AM
To: 'nmusers_at_globomaxnm.com'
Subject: [NMusers] nonmem question regarding condition number

Dear nmusers,
I am currently analyzing PK profiles after a single iv adminstration. Volun=
teers were sampled frequently: 12 samples per curve.
For the structural model 2- and a 3-compartment models are compared. The CW=
RES vs time plot of the 2-compartment shows the typical wave, indicative fo=
r a 3-compartment model. Introduction of a third compartment eliminated the=
 wave, OFV decreased with 180 points, precision of the estimates did change=
 notably, but the condition number increased from 998 to 2.0*E16.
Does anyone know how to interpret this large increase in condition number?
Best wishes,
Laureen ten Berg

[Beschrijving: cid:image001.gif_at_01CD804B.C9EE1BA0]




Laureen ten Berg-Lammers

    Hospital Pharmacist


Academisch Medisch Centrum
Universiteit van Amsterdam




Meibergdreef 9 | 1105 AZ Amsterdam | Room: M01-224
Tel: +31(0)20-5662726 | Pager *62726 | l.a.tenberg-lammers_at_uva.amc.nl<mailt=
o:l.a.tenberg-lammers_at_uva.amc.nl>



________________________________

AMC Disclaimer : https://www.amc.nl/disclaimer

________________________________


image001.gif
(image/gif attachment: image001.gif)

Received on Thu Mar 24 2016 - 09:22:40 EDT

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to: nmusers-request@iconplc.com. Once subscribed, you may contribute to the discussion by emailing: nmusers@globomaxnm.com.