10 replies to this topic

[jprell]

Dear all,

 

what is the difference between Predicted concentrations on Summary Table sheet and Predicted data sheet for PK model output? There is large difference between them and only Predicted data sheet is used for Predicted Y vs Observed Y vs Time plots.

 

Also, item CORR_(OBS,PRED) on Diagnotics sheet is related to which predicted concentrations?

 

Thanks, Jan.

 

[bwendt@certara.com]

Hi Jan,

 

I see that you are using the legacy PK models in Phoenix. The main difference between the Summary Table and the Predicted data sheet is the resolution of time, e.g. x-axis. The Summary table will only write out predicted concentrations for the time points of your observations, e.g. your input. The Predicted data sheet is for plotting purposes, to give a smoother line when overlayed with the observed concentrations in the Observe Y and Predicted Y vs X plot.

 

The correlation in diagnostics sheet is using the observed concentrations (Obs) and the predicted concentrations (PRED). 

 

Hope, this is clear. Let me know if there are further questions.

 

Bernd

[jprell]

Hi Bernd,

 

thanks for your answer!

 

"The Summary table will only write out predicted concentrations for the time points of your observations, e.g. your input."

Understand. But what I do not understand is the reason why the early part (i.e., 0.5 - 2 hrs) of the PK profile is skipped in Predicted data?

 

 

 

"The correlation in diagnostics sheet is using the observed concentrations (Obs) and the predicted concentrations (PRED)."

I am sorry, my question was not clear. I meant whether predicted concentration from Summary Table sheets or from Predicted data sheet are used for correlation coefficient with observed data?

 

Thanks and have a nice day, Jan.

Edited by jprell, 11 May 2023 - 06:12 AM.

[bwendt@certara.com]

Hi Jan,

 

the early part is not skipped in Predicted Data, the time slots are spread evenly over the entire time range and always uses 1,000 points. You can see that the distances between timepoints are always the same.

 

The value for correlation between observed and predicted data comes straight out of the engine but it can only compare data when you have values for both observed and predicted concentration, so it is essentially using the information from the Predicted Data worksheet.

 

Bernd

[jprell]

Hi Bernd,

 

thanks for your answer!

So I can get more detailed early part with increased number of predicted points and/or with cutting the profile, let's say, after 504 hrs, am I right?

 

Thanks and have a nice day, Jan.

[bwendt@certara.com]

Jan,

 

I would not cut the profile, but you can specify more points for the predicted data under engine settings:

 

 

Alternatively, I would recommend switching to the modern Phoenix Model where you can specify all of these kind of details much more elegantly. Let me know when you are interested in that and I can provide more help.

 

Bernd

[jprell]

Hi Bernd,

 

thanks a lot! and yes, I am interested in Phoenix Model.

 

Have a nice day, Jan.

[bwendt@certara.com]

Hi Jan,

 

we have run a series of webinars covering a range of modeling scenarios where the Phoenix Model object was used:

 

Phoenix Modeling Language (PML) School | Certara

 

There are recorded sessions and we do have all the materials available in this forum at:

 

 

PML School - Certara Forums

 

That should give you a great start.

 

Bernd

[jprell]

Hi Bernd,

 

sorry for the late answer. Thanks for the provided links - which WNL module should I have to run Phoenix Models?

 

Thanks, Jan.

[Simon Davis]

Base WNL license gives you access to the Phoenix models where you can make use of most features but if you want to model Random Effects you would need NLME license (i.e. for population analyses)

[jprell]

Dear Simon,

 

thanks, I will look into it.

 

Have a nice day, Jan.