NONMEM Users Network Archive

Hosted by Cognigen

Re: Stepwise covariate modeling

From: Sebastien Bihorel <Sebastien.Bihorel>
Date: Tue, 29 Oct 2019 14:43:27 +0000


I am not 100% certain that a decrease in the IIV terms associated with the =
covariate-parameter relationships is a criteria used in the automated selec=
tion process implemented by the PsN scm command (at least I could not find =
any reference in the documentation). You may want to implement a more manua=
l approach to the problem where you define your own set of criteria for cov=
ariate selection that you would apply at each step.

PS: our KIWI platform can help you with the automated creation of univariat=
e runs at each step and summarization of results while keeping you in contr=
ol of the covariate selection criteria...

Sébastien Bihorel
Director, Pharmacometrics and KIWI™ applications
Cognigen Corporation, a SimulationsPlus company
Buffalo Office: +1 716 633 3463 ext. 323 | Website<https://www.simulations-=>
From: owner-nmusers
 of Singla, Sumeet K <sumeet-singla
Sent: Tuesday, October 29, 2019 10:00
To: nmusers
Subject: [NMusers] Stepwise covariate modeling


I am performing stepwise covariate modeling using PsN feature in Pirana. I =
am getting some covariates which are statistically reducing OFV significant=
ly, however, when I include those covariates in the PK model, the results I=
 am getting are exactly similar to what I am getting in my base model, i.e.=
 there is no difference in individual predictions or pop predictions or any=
 other diagnostic plots. So, does that mean I should move forward WITHOUT i=
ncluding those covariates as they don’t seem to be explaining inter-indiv=
idual variability despite scm telling me that they are statistically signif=


Sumeet K. Singla

Ph.D. Candidate

Division of Pharmaceutics and Translational Therapeutics

College of Pharmacy | University of Iowa

Iowa City, Iowa



Received on Tue Oct 29 2019 - 10:43:27 EDT

The NONMEM Users Network is maintained by ICON plc. Requests to subscribe to the network should be sent to:

Once subscribed, you may contribute to the discussion by emailing: