Re: AW: [NMusers] $Error log transformed

From: Nick Holford <n.holford_at_auckland.ac.nz>
Date: Mon, 19 Dec 2016 18:15:02 +1300

Katrin,

It is not good practice to add a small quantity to the prediction to
avoid numerical errors. Stuart Beal (the original developer of NONMEM)
wrote this comment in 2004:

"..., this will very often alter the results, especially when during the Estimation Step parameter
values are considered that lead to smaller values of F than might be imagined."

http://www.cognigencorp.com/nonmem/nm/98feb112004.html

He supplies some workarounds which you may wish to consider.

You may also wish to use a more realistic error model that includes an
additive component that would be appropriate for a concentration
approaching zero. All real measurements will have some noise when the
true value is zero.

Best wishes,

Nick

On 18-Dec-16 07:46, Katrin Volz wrote:
>
> Hey Rob,
>
> Thank a lot. It seems to work well.
>
> Im quite happy with this log transformation :)
>
> And yes, thanks for clarifying that I have to take the natural logarithm.
>
> Cheers
>
> Katrin
>
> *Von:*owner-nmusers_at_globomaxnm.com
> [mailto:owner-nmusers_at_globomaxnm.com] *Im Auftrag von
> *R.terHeine_at_radboudumc.nl
> *Gesendet:* 16 December 2016 11:24
> *An:* ankekatrin.volz_at_uni-saarland.de; nmusers_at_globomaxnm.com
> *Betreff:* RE: [NMusers] $Error log transformed
>
> Hi Katrin,
>
> Youll have to take the natural logarithm of your DVs
>
> I would change your error block to:
>
> $ERROR
>
> IPRED=LOG(F+1E-16) ; Protection against LOG(0) in case of dosing
> events etc
>
> Y=IPRED+ERR(1) ; FOR PROPORTIAL ERROR
>
> Cheers!
>
> Rob
>
> *Van:*owner-nmusers_at_globomaxnm.com
> <mailto:owner-nmusers_at_globomaxnm.com>
> [mailto:owner-nmusers_at_globomaxnm.com] *Namens *Katrin Volz
> *Verzonden:* vrijdag 16 december 2016 10:35
> *Aan:* nmusers_at_globomaxnm.com <mailto:nmusers_at_globomaxnm.com>
> *Onderwerp:* [NMusers] $Error log transformed
>
> Dear All,
>
> I would like to model my data log transformed.
>
> So far, I already found some suggestions for this problem here but
> none of them worked in my case - probably I did sth. wrong.
>
> I would be very happy if someone could help me with changing my error
> model, which I assume has to be changed.
>
> (I also assume that the dataset has to be changed from DV to LOG(DV)?)
>
> So, here is the code I used so far (not log transformed):
>
> $ERROR
>
> IPRED = F ;IPRED drug in the central
> compartment
>
> IF (CMT.EQ.5) IPRED = A(5) ;IPRED second drug in 5th compartment
>
> DEL=0
>
> IF (IPRED.EQ.0) DEL=0.0001
>
> W = F
>
> IRES = DV - IPRED
>
> IWRES = IRES /(W + DEL)
>
> Y = IPRED + IPRED * EPS(1) ; error
> drug in the central compartment
>
> IF (CMT.EQ.5) Y = IPRED + IPRED * EPS(2) ; error second drug in
> 5th compartment
>
> It would be great to get some help.
>
> Thanks!
>
> Katrin
>
> Het Radboudumc staat geregistreerd bij de Kamer van Koophandel in het
> handelsregister onder nummer 41055629.
> The Radboud university medical center is listed in the Commercial
> Register of the Chamber of Commerce under file number 41055629.
>

--
Nick Holford, Professor Clinical Pharmacology
Dept Pharmacology & Clinical Pharmacology, Bldg 503 Room 302A
University of Auckland,85 Park Rd,Private Bag 92019,Auckland,New Zealand
office:+64(9)923-6730 mobile:NZ+64(21)46 23 53 FR+33(6)62 32 46 72
email: n.holford_at_auckland.ac.nz
http://holford.fmhs.auckland.ac.nz/
http://orcid.org/0000-0002-4031-2514
Read the question, answer the question, attempt all questions

Received on Mon Dec 19 2016 - 00:15:02 EST

This archive was generated by hypermail 2.3.0 : Fri Sep 27 2019 - 16:53:05 EDT