Jump to content


Photo

proportional error parameter (tvCMultStd) as a negative value?


  • Please log in to reply
2 replies to this topic

#1 Davidq8

Davidq8

    Member

  • Members
  • PipPip
  • 19 posts

Posted 07 December 2021 - 12:08 AM

Hi 
 
Is it possible to estimate the proportional error parameter (tvCMultStd) as a negative value?
I am trying to build a population Pk model using a combined additive and proportional error model; however, I got tvCMultStd as a negative value. Interestingly, when I use only a proportional error model, it is estimated as a positive value. I want to use a combined additive and proportional error since it shows a much better fit confirmed by AIC and diagnostic plots than the proportional error model. Still, I have a concern regarding the negative estimation of tvCMultStd. 
 
Can I get any explanation regarding this?
 
Thanks


#2 Ana Henry

Ana Henry

    Advanced Member

  • Members
  • PipPipPip
  • 227 posts

Posted 07 December 2021 - 12:34 AM

Dear David,

You are correct and tvCMultdStd should not be negative.  However, in Phoenix, tvMultStdev is treated as a regular fixed effect and thus the reason it could become negative.  In training, we always recommend that a lower bound of 0 is added to this parameter.

 

Hope this helps,

Ana Henry

 

 

 

Hi 
 
Is it possible to estimate the proportional error parameter (tvCMultStd) as a negative value?
I am trying to build a population Pk model using a combined additive and proportional error model; however, I got tvCMultStd as a negative value. Interestingly, when I use only a proportional error model, it is estimated as a positive value. I want to use a combined additive and proportional error since it shows a much better fit confirmed by AIC and diagnostic plots than the proportional error model. Still, I have a concern regarding the negative estimation of tvCMultStd. 
 
Can I get any explanation regarding this?
 
Thanks

 



#3 Davidq8

Davidq8

    Member

  • Members
  • PipPip
  • 19 posts

Posted 11 February 2022 - 08:22 PM

Dear David,

You are correct and tvCMultdStd should not be negative.  However, in Phoenix, tvMultStdev is treated as a regular fixed effect and thus the reason it could become negative.  In training, we always recommend that a lower bound of 0 is added to this parameter.

 

Hope this helps,

Ana Henry

Dear Ana

Thanks for your help. Your suggestion has fixed the issue!






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users