Model results Phx 6.4 vs 7.0

Phoenix Certara Forums Modelling and Simulation
1

Dear Certara Team,

we were running a PK/PD model using QRPEM under Phx 6.4 (better to say NMLE 1.3). We then created an exact copy of the phxproj file and re-ran it using Phx/NLME 7.0, also QRPEM. We now see that 1. the thetas (some more, some less) and stdevs are different and 2. The CWRES plot in Phx7.0 suddenly blows up (see the y-scale in the attached document). The CWRES plot using Phx 6.4 is “normal” (i.e. most data within the +/-2 bounds).

Could you kindly help us here?

Thanks

Frank

Phx 6.4 vs 7.0.pdf (259 KB)

2

Difficult to know without seeing the project. Can you try increase the number of smaples to 3000 and check the option MAP.

QRPEM will sample every time different samples and it can lead to different results.

Now, not usre if your model converged on both cases.

Then increasing number of smaples to 3000 and using MAP option can lead to better convergence nd also convergence between both runs.

best

Serge

3

Thanks Serge for the reply. Using more samples and the MAP assist helps stabilizing the fit.

But still I do get that strange Y scale in the CWRES plot, i.e. the CWRES are very high with Phx7.0, although otherwsie the fit looks OK! See pdf above in my initial post.

If I comment out (#) the PD model and simply fit the PK, the CWRES is fine (i.e most between +/-2). If I put back the PD, the PK params do not change but suddenly the CWRES for PK blows up.

What could be the problem here?

4

Hi Frank, would it be possible to send you project to certara support ?

can you share the observe statements from your model

observe for conc and for pd

Thanks !

Samer