Jump to content


Photo

IWRES and NPDE vs time/concentration plots interpretation

Modelling interpretation IWRES NPDE

  • Please log in to reply
2 replies to this topic

#1 jmt3

jmt3

    Newbie

  • Members
  • Pip
  • 5 posts

Posted 03 April 2022 - 06:21 AM

https://www.linkpict...m/q/Model-1.png

https://www.linkpict...m/q/Model-2.png

https://www.linkpict...m/q/Model-3.png

 

Hello everyone,

I have a question with regards to the scatter plot of residual (especially IWRES and NPDE vs time/concentration). I have two different plots here simulating from two different structural models in NLME PopPk model (see attached Model 1 and Model 2 plots).

My understanding is that the scatter plot of residual tells you how much the predicted data deviates from the observed data. Based on these two plots (IWRES and NPDE of model 1 vs model 2), my understanding is that Model 1 is better than Model 2 only because the interpolation spline of Model 1 (the yellow line) lies on the horizontal y=0 line. In contrast, Model 2 has the trend line scattering unevenly on the horizontal line, and it even increasingly shows deviation when the time and the concentration are higher. Is my interpretation correct?

How about the plot from Model 3 (see attached)? I won't consider this model because all the data skewed towards right, and hence it is telling model misspecification? am I correct in this?

 

I hope this makes sense.

 

Thanks in advance.


Edited by jmt3, 03 April 2022 - 06:22 AM.


#2 smouksassi1

smouksassi1

    Advanced Member

  • Members
  • PipPipPip
  • 231 posts
  • LocationMontreal

Posted 04 April 2022 - 09:03 AM

Hi,

It is hard to help you without details and at least what are you trying to do ?

need to see Response and Cc versus time and some about the design of your study ?

What are the two version of the structural model ?

 

you seem to have some structural model deficiencies at the second time point ( or group of time points)

yes you want a flat line  but from your plots hard to compare with different scales and without knowing what changed.

might need to log transform your Cc if you have multiple log values


  • jmt3 likes this

#3 jmt3

jmt3

    Newbie

  • Members
  • Pip
  • 5 posts

Posted 04 April 2022 - 09:23 AM

Hi,

It is hard to help you without details and at least what are you trying to do ?

need to see Response and Cc versus time and some about the design of your study ?

What are the two version of the structural model ?

 

you seem to have some structural model deficiencies at the second time point ( or group of time points)

yes you want a flat line  but from your plots hard to compare with different scales and without knowing what changed.

might need to log transform your Cc if you have multiple log values

Hi Smouksassi,

Thank you for your reply. The plots have been log-transformed both in time (top plot) and Cc (bottom plot). I just need to know how to read the plots. Based on my interpretation, I think the plot from Model 1 is the best among all three, but I wonder what is wrong with Model 2 and Model 3. Could you please tell me what is the reason that makes the line tilted down to the right in Model 2 and the line skewed to the right in Model 3?







Also tagged with one or more of these keywords: Modelling, interpretation IWRES NPDE

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users