Thank you for organizing - I have written multiple model in the old WNL5 ASCII format, and appreciate the opportunity to learn to use the newer language. This morning’s session was excellent.
Thank you for providing these training sessions / webinars. The morning session was very good. Would it be possible to provide the dataset used for the modeling / presentation?
unfortunately we cannot provide the datasets. This was part of the arrangement with the publisher of the book. However, as you have seen in the webinar, the datasets are usually pretty generic, single profile type of datasets with just a few columns like time and concentration.
Could you confirm that PML is used in the excercises/case studies provided the 5th edition of the book from J. Gabrielsson and D. Weiner (rather than WinNonlin modeling language in the 4th Ed.)?
yes, Johan and Dan have changed the language from the legacy WNL ASCII code to PML. All exercises in the 5th edition are provided as Phoenix project files on a USB-stick that comes with the book. There are more than 100 exercises in the book by now.
I am glad everyone found yesterday’s session useful, just a reminder of how to stay informed on these forums; Want to be notified of all new topics posted to a forum?
Click the “Follow this forum” button at the top right of the main page for each forum that interests you.
Want to be notified of all replies to a topic of interest?
Click the “Follow this topic” button at the top right of that topic’s page.
(if you don’t do the above you won’t receive emails regarding new topics or new posts)
Where is the link to the recorded video? Thank you.
I’d like to thank you for all your effect to teach PML. I was looking for PML tutoring for a long time. I greatly appreciate your team’s kindness. I’ll follow all your 20 PML live webinars.
It takes a little longer to upload our recorded webinar to the Certara Youtube channel. In the meantime, if you want to watch the recording, please use the link below:
Greetings. Thanks for the great organization and kickoff session yesterday on PML, very helpful.
For the series of sessions, will they be scheduled for the same time as the last one (i.e. 10am EST on Thursday every two weeks)? I am hoping to block my calendar so that I have a better chance to attend to these seminars as much as possible, and will appreciate your input on the timing.
Thanks all for organizing this series and it was a great learning resource. I submitted a question at the end of webinar and received a reply thereafter from Chris, shown as follows. Thanks for taking care of the question.
Q: In the structural parameters, we only have CL, but in the fixed effects, we have a, Cl0 and CL, why?
A: In the non-static clearance model, Cl is defined as a function of the fixed effects Cl0 and a. It is not a parameter in itself – it is derived from other parameters.
But still I am little bit confused. When we define Cl = Cl0 - a*C, Cl0 and a are constant parameters and Cl is a variable parameter.
Do we have to use stparm to define Cl0 and a as parameters?
Is “fixef(tvCl = c(, 4.3, ))” necessary? Cl is dependent on C and therefore changing over time. Should we give it a population mean?
So I am wondering how the model would work with these codes.
test(){
deriv(A1 = - Cl * C - Cl2 * (C - C2) - Cl3 * (C - C3))
deriv(A2 = Cl2 * (C - C2))
deriv(A3 = Cl3 * (C - C3))
dosepoint(A1)
C = A1 / V
C2 = A2 / V2
C3 = A3 / V3
Cl = Cl0 - a * C
error(CEps = 1)
observe(CObs = C * (1 + CEps))
stparm(V = tvV)
stparm(V2 = tvV2)
stparm(V3 = tvV3)
stparm(Cl = tvCl) stparm(Cl0 = tvCl0) ## what if we add this? stparm(a = tva)## what if we add this?
stparm(Cl2 = tvCl2)
stparm(Cl3 = tvCl3)
fixef(a = c(, 0.0005, ))
fixef(Cl0 = c(, 4.3, ))
fixef(tvV = c(, 0.5, ))
fixef(tvV2 = c(, 1.7, ))
fixef(tvV3 = c(, 1.7, )) fixef(tvCl = c(, 4.3, )) ## what if we remove this?
fixef(tvCl2 = c(, 6.5, ))
fixef(tvCl3 = c(, 2.2, ))
secondary(Min_Cl=Cl0 - a * 1000)
}
thanks for your feedback. Just to clarify the confusion around constant parameters, variable parameters and derived parameters:
Cl is a derived parameter, it is not involved in the fitting process, it is estimated on the basis of the variable parameters a and Cl0
both Cl0 and a are variable parameters. Their values will change through the fitting process, the starting value is shown under the fixed effects section of the textual model:
fixef(a = c(, 0.0005, ))
fixef(Cl0 = c(, 4.3, )
Any time, the value for a and Cl0 will change, the value for Cl will also change, since Cl is derived from a and Cl0 through the equation:
Cl = Cl0 - a * C
There are no constant parameters in this model.
Hope, this helps. Let me know if there are more questions.
Thanks so much for the quick clarification. I get you and totally agree. My point was Cl0 and a would each have a constant as population mean but Cl wouldn’t. Right?
If so, my 2 questions seem to still exist…Appreciate any more advice.
sorry, it is getting late my time, I was not clear. We have not discussed the population scenario, we have just discussed the individual fitting scenario.
Population PK is a more difficult scenario and is not very well suited for a beginner’s class. But to take your point, if we would have more than one profile to fit, assuming the different profiles originate from different subjects, then we will switch to population mode, in which case we will fit the data to find a typical population estimate for both a and Cl0. At the end of the fitting process you will have indeed just a single estimate for a and Cl0 resp.But there will be certain random effects as well, that is, profiles between the subjects will show differences and these differences are covered by what we call etas (random effect parameters).
Thanks again and yes your answer is helpful. But my question was actually not on population model or not. I don’t know if this forum allows lengthy discussion, but if not violating the forum policy I don’t mind to bring up all my dump questions. :rolleyes:
Let me put my questions in a different way.
Why Cl0 and a don’t need to be predefined using stparm while all other parameters need to?
It seem tvCl is not reported (slide 11). Do we really need to have the code of fixef (tvCl = c (, 4.3, ))? I am assuming fixed effect parameters should not be dependent on time or concentration? If so, the concentration-dependent Cl seems not to be a fixed effect parameter.
Unfortunately I don’t have the source data otherwise I’d test if my codes would generate any errors or result in same estimates.
there are no structural parameters defined for a and Cl0
the fixed effects are defined for a (0.0005) and Cl0 (4.3) but not for Cl
you don’t see Cl reported in the Theta table where all the fixed effects are reported
Re 1.: I guess that Dan just want to keep the changes to the textual model as small as possible, therefore he omitted to add a and Cl0 as structural parameters. In general , you would define those additional parameters as structural parameters in a population model, since this would allow you to add random effects. Since we are dealing with an individual fit in this particular exercise, this step is unnecessary.
Re 2.: this is different from the code that you posted. The variable parameters a and Cl0 need to have initial estimates, which are defined here. You don’t want an initial estimate for Cl since this is a derived parameter based on a and Cl0
Re 3.: There is no single estimate for clearance since it is dependent on the concentration. If you want to see the variability of Cl you would need to add another table under the Run Options of your Phoenix Model that will output C and Cl versus time (see picture attached).
