From: "Gordi, Toufigh" Toufigh.Gordi@cvt.com
Subject: [NMusers] WRES vs. time 
Date: Wed, September 1, 2004 7:20 pm

Dear all,

Using Xpose, I have a plot of WRES vs. TIME, which indicates almost all values are
larger than 0 at later time-points. Could anybody point out to me what such a trend
depends on and how I can remedy it. In general, the individual fits are quite ok and
the WRES values are quite small but it is somewhat bothering that there is a trend.

Best regards,

Toufigh Gordi

_______________________________________________________

From: drfreedman@drfreedmaninc.com
Subject: RE: [NMusers] WRES vs. time
Date: Wed, September 1, 2004 7:55 pm 

Toufigh

Do you need a disease progression model?

Immanuel Freedman, PhD, MIEEE
(619) 884-1347
_______________________________________________________

From: "Wang, Bing" Bing.Wang@Abgenix.com
Subject: RE: [NMusers] WRES vs. time
Date: Wed, September 1, 2004 8:01 pm

I wonder if you have many BLQ observations at the end of the profiles. These BLQs
might explain the apparent bias since they are usually excluded in the analysis...
Just a thought. 
_______________________________________________________

From:"bvatul" bvatul@verizon.net
Subject: RE: [NMusers] WRES vs. time
Date: Wed, September 1, 2004 8:36 pm

Hello Toufigh

It is likely that your model is not able to describe the data well at later
time points. You can include more complexity to your model (One vs Two
compartment etc) and would probably help you.

Venkatesh Atul Bhattaram
CDER, FDA.
_______________________________________________________

From: "Mats Karlsson" mats.karlsson@farmbio.uu.se
Subject: RE: [NMusers] WRES vs. time
Date: Thu, September 2, 2004 1:18 am

Toufigh,

I agree with what has been suggested before. Just one additional point.
There may be nothing wrong with your model but with you diagnostic plot.

WRES is the "perfect" residual to inspect when we use the first-order
(FO) method as there is a direct correspondence between the estimation
method and the residual. However, not so when we use conditional
estimation methods (FOCE, FOCE INTER), but we still use it lacking
anything better. We often take it for granted that diagnostic plots
should show no pattern (or a particular pattern similar for all
applications). This is often not the case. I suggest that you take you
final parameter estimates and your study design (realized doses and
sampling times), simulate new data and obtained predictions and WRES (in
one go using $SIML and $EST MAXEVAL=0). If the pattern then is similar
to what you saw with real data, I would be satisfied that the WRES
pattern is no concern.

Best regards,
Mats

--
Mats Karlsson, PhD
Professor of Pharmacometrics
Div. of Pharmacokinetics and Drug Therapy
Dept. of Pharmaceutical Biosciences
Faculty of Pharmacy
Uppsala University
Box 591
SE-751 24 Uppsala
Sweden
phone +46 18 471 4105
fax   +46 18 471 4003
mats.karlsson@farmbio.uu.se
_______________________________________________________