Jump to content


Photo

out of range final -LL value 1.0000000001E+070


  • Please log in to reply
3 replies to this topic

#1 joybaker

joybaker

    Member

  • Members
  • PipPip
  • 13 posts

Posted 14 April 2021 - 02:25 AM

I wonder if anyone have come across similar problem like me?

 

This information always appears after I build up my model:

Model execution failed

"out of range final -LL value 1.0000000001E+070"

 

I use FOCE-ELS, and this is the model that have been developed with good fixed parameter and good AIC. And I only change a little bit of the initial parameters (say change 30 to 35), or adding covariate, so I don't think this is because I didn't choose a good set of initial values.

 

The other thing is that in the NLME Job Status TAB, I can see the the 2LL value is going down and close to what I usually get. So the "out of range final -LL value" information doesn't make any sense to me.

 

 I have changed the algorithm to naive pooled or other method and change it back to FOCE-ELS, but this message still came up and the calculation stoped.

 

So any suggestion for this problem, should I try to adjust parameters in the running mode?

Attached Thumbnails

  • 0210414101702.png
  • 120210414102426.jpg


#2 cradhakr

cradhakr

    Member

  • Members
  • PipPip
  • 17 posts

Posted 14 April 2021 - 02:55 AM

Hi Joy,

 

Is it possible for you to share the project file? If you dont want to share it on the forum you could email it to support@certara.com



#3 smouksassi1

smouksassi1

    Advanced Member

  • Members
  • PipPipPip
  • 179 posts
  • LocationMontreal

Posted 14 April 2021 - 05:55 AM

do you have a proportional error model with super small prediction ? try to add a dummy additive error that you fix to a small value if you do not need it it is not uncommon to get near zero or numerically zero prediction with lag time models did you try a transit abs model ?



#4 joybaker

joybaker

    Member

  • Members
  • PipPip
  • 13 posts

Posted 06 May 2021 - 08:12 AM

do you have a proportional error model with super small prediction ? try to add a dummy additive error that you fix to a small value if you do not need it it is not uncommon to get near zero or numerically zero prediction with lag time models did you try a transit abs model ?

Thank you. I think your suggestion dose work. After the fitting, it just give a too small additive error to the final fitting so it can't work normally even if I just ran it again when I accept those values.

 

I do have following questions I sent it through the email related with bootstrap results, hope you can help thanks!






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users