Jump to content


Photo

PopPK VPC not including all observed data points

vpc poppk

  • Please log in to reply
5 replies to this topic

#1 cardtrick12

cardtrick12

    Newbie

  • Members
  • Pip
  • 3 posts

Posted 20 April 2022 - 04:07 PM

Dear everyone,

 

I am working through a VPC and I am trying to figure out why the shading is not extended all the way to the end of the observed data points?

 

I've attached the image here of what I am seeing.

 

I have tried adjusting the binning manually but it doesn't seem to work. Any thoughts? Happy to provide additional information as needed.

 

Thanks!

Attached Thumbnails

  • VPC_04202022.png

Edited by cardtrick12, 20 April 2022 - 04:37 PM.

  • Naitoeste likes this

#2 smouksassi1

smouksassi1

    Advanced Member

  • Members
  • PipPipPip
  • 231 posts
  • LocationMontreal

Posted 20 April 2022 - 06:29 PM

are you using kmeans or other binning what happens you dont use binning at all ?

it seems you have mutliple doses are you doing pred correction and why not do the vpc versus TAD.

 

I typically use tidyvpc to do more advanced vpc plots



#3 cardtrick12

cardtrick12

    Newbie

  • Members
  • Pip
  • 3 posts

Posted 24 April 2022 - 05:29 PM

Using no binning, we get the attached image, so we do get the inclusion of the data points at the end, although the predictions overlap with one another.

 

Might just be easier if I attach the project file, which is also attached. The work with VPCs is under the 'TFVdp Model' Workflow under 'POP ML Model 1_Cov_VPC' (second from bottom).

 

I'd appreciate any help you could cordially provide. I'll take a closer look into tidyvpc.

 

 

Edit: momentarily will attach Phoenix project file; receving an error

 

Edit #2: Couldn't attach the Phoenix file but it is here via DropBox: https://www.dropbox....22.phxproj?dl=0

Attached Thumbnails

  • VPC_nobinning_04202022.png

Edited by cardtrick12, 24 April 2022 - 05:35 PM.


#4 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,328 posts

Posted 28 April 2022 - 11:10 AM

As Samer said, in some situations, K-means binning does not work well - in this case we can propose explicit centers option, (or to go to tidyVPC).

 

  Simon.

 

Attached Thumbnails

  • explicit_centres.png


#5 cardtrick12

cardtrick12

    Newbie

  • Members
  • Pip
  • 3 posts

Posted 28 April 2022 - 12:51 PM

As Samer said, in some situations, K-means binning does not work well - in this case we can propose explicit centers option, (or to go to tidyVPC).

 

  Simon.

 

Thanks Simon! I'll explore the explicit centers option. From your image, could you tell me what the centers were that you used? I'd like to re-create the image you've shown. Thanks!



#6 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,328 posts

Posted 28 April 2022 - 01:16 PM

You can see most of them in the screen shot at the bottom;

0,24,48,72,168,264,336,528,696,864







Also tagged with one or more of these keywords: vpc, poppk

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users