Since theoretically rescaling should not matter, the problem is in details that can not be extracted
from the description without seeing the data and control stream. Just as an experiment (and to check
your data creation process) run the model with the data where AMT and DV are rescaled by a factor
of, say 1.1. If you see the difference there, you probably made an error in the data creation
script. If it runs OK (returns exactly the same parameter estimates), increase (or decrease) the
rescaling factor until it break the solution. Let us know what happens, it may help to diagnose the
problem. If the problem appear at some value of the rescaling factor, it may indicate rounding error
or something similar. Try to use mcmol or nmol or anything in between to have concentration and dose
of the same order of magnitude as in mg form.
Another test: try to start with the solution as initial values.
Yet another try: use simpler 1-compartment (ADVAN2) model for both sets of data to see whether the
problem is universal or ADVAN-specific. If it is universal, it would point to the data creation error.
Leonid
[EMAIL PROTECTED] wrote:
Nick,
oh, sorry about that, that was a typo. I have mg and µg/L and µmol and
nmol/L, so I'm afraid that's not the reason. I have a combined error
model. However, although I needed the additive error to get the model to
run, I fixed it to a very small value (0.0001). But to make sure I just
adjusted this value too, with the same result.
Just to get this clear for me...in theory the parameters should be the
same, shouldn't they? So if this keeps on happening, what else can I do? I
was thinking about recalculating the metabolite concentrations by dividing
them with their molar mass and then multiplying with the molar mass of
the parent. Would this be a reasonable approach, just to pretend that the
metabolite is parent by using this calculation? Would the metabolite
parameters still be correct then?
Thanks for your help.
Nele
_________________________
Nele Plock
Bayer Schering Pharma AG
GPD/Pharmacokinetics
Metabolism & Bioanalysis
D- 13342 Berlin
Phone : +49-30-468 15146
Fax: +49-30-468 95146
[EMAIL PROTECTED]
Nick Holford <[EMAIL PROTECTED]>
08.03.2007 15:10
An
[EMAIL PROTECTED]
Kopie
[email protected]
Thema
Re: [NMusers] change from mass to molar units
Nele,
Two suggestions:
1. You have mass units of mg and mg/L but molar units of nmol/1000 and
nmol/L. So the ratio of AMT to DV is different by a factor of 1000
2. If you have an additive error model then the scale of the residual
error model parameter will need to change.
Best wishes,
Nick
[EMAIL PROTECTED] wrote:
Dear all,
I have fitted a simple two-compartment model with linear elimination to
some compound data (ADVAN4 TRANS4). Everything worked well. Because I
want
to extend this model to account for metabolite data as well, I now
wanted
to change all data to molar units, i.e. I changed the AMT and DV column
in
the same way. To my understanding, if I change both these columns, the
obtained parameters should end up being the same. However, when I use mg
and µg/mL the model runs just fine, whereas when I use µmol and nmol/L
the
parameter estimates change completely and always run into some boundary.
Can anyone shed some light on why this happens? Is there any way to get
around this? Could I keep the mass units and do some changes in the
control stream when incorporating the metabolite data? Any help would be
highly appreciated.
Best regards
Nele
_________________________
Nele Plock
Bayer Schering Pharma AG
GPD/Pharmacokinetics
Metabolism & Bioanalysis
D- 13342 Berlin
Phone : +49-30-468 15146
Fax: +49-30-468 95146
[EMAIL PROTECTED]
--
Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
Zealand
email:[EMAIL PROTECTED] tel:+64(9)373-7599x86730 fax:373-7556
http://www.health.auckland.ac.nz/pharmacology/staff/nholford/
