2 replies to this topic

[Cathrine Denton]

I am working on a PK model for data from various tissues. I fit a simple 2CM PK model to the plasma data using 1/Yhat weighting. However, when I build in a compartment (using ASCII model) to model liver concentrations, I don't know how to keep the 1/Yhat weighting for plasma, but use a different weight (say 1/Y) for the liver data. Is there a way to do this?

[Jun Shen]

The WNL manual mentions a way by defining the weighting variable WT,

 

FUNC 1

F=....

WT=1/(F*F)

END

 

FUNC 2

F=....

WT=1/Y

END

 

I never tried myself, will be interested to know if this is working.

 

Jun Shen

Seventh Wave Labs

[Simon Davis]

Sorry - a long time in coming to this response but I would suggest this is much easier in the Phoenix models; e.g. here is a snippet of code from a simple PKPD link model where different error models have been applied in the user interface to the conc and effect of multiplicative and additive respectively;

/extranet/media/kunena/attachments/legacy/images/errDiff.jpg

to give

error(EEps = 1)
observe(EObs = E + EEps)
error(CEps = 1)
observe(CObs = C * (1 + CEps))

but yes the different functions you described above should work in WNL classic ascii models

Simon