From: zhang.239@osu.edu
Subject: [NMusers] 1 simple question from a nmbeginner
Date: 10/12/2003 11:43 PM

Dear all,

I have 1 simple question:

From NONMEM menu, I know that a decrease in OFV higher than 3.8 was required to
incorporate a covariate into the model.

In one data set, I found that the OFV decreased from -320 to -330 after
incorporated BSA to the model. Can anybody let me know how important this change (3%) is?

Thanks!

Yilong Zhang
_______________________________________________________

From: bachmanw@globomax.com
Subject: RE: [NMusers] 1 simple question from a nmbeginner
Date: 10/13/2003 8:56 AM

The % change is meaningless.  The criterion for inclusion/exclusion is based on your
chosen signifigance level, the degrees of freedom and the critical chi square value
(for sig. level and df).  Also, change in objective function should not be the only
criterion for acceptance.  You should also consider the change in variance estimates
and the diagnostic plots before accepting the new model.
_______________________________________________________