Error Resul Full Latent Choice Model

52 views
Skip to first unread message

azis hakim

unread,
Nov 17, 2022, 8:33:00 AM11/17/22
to Biogeme
Dear Michel,

I am trying run Full latent Choice Model..
However the result not really satisfied... 
There are error number in the result such as : 1.8e+308. Do you have some advice to fix this ?
The example of error can be seen in following picture..error result latent choice model.jfiferror result latent choice model-1.jfif  

Bierlaire Michel

unread,
Nov 17, 2022, 9:01:39 AM11/17/22
to azis....@gmail.com, Bierlaire Michel, Biogeme
These errors are reported only in the section about second order information (correlation, covariance, etc.) It does that when trying to calculate square roots of negative numbers. A. probably cause is that the algorithm has not fully converged, and that the second derivative matrix is not positive definite at the value of the parameters.

Which algorithm are you using? 
Did you try a sequential estimation first? If so, use the estimated parameters of the sequential estimation as starting points for the full information estimation. 



On 17 Nov 2022, at 13:50, azis hakim <azis....@gmail.com> wrote:

Dear Michel,

I am trying run Full latent Choice Model..
However the result not really satisfied... 
There are error number in the result such as : 1.8e+308. Do you have some advice to fix this ?
The example of error can be seen in following picture..<error result latent choice model.jfif><error result latent choice model-1.jfif>  


--
You received this message because you are subscribed to the Google Groups "Biogeme" group.
To unsubscribe from this group and stop receiving emails from it, send an email to biogeme+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/biogeme/e9ceb685-aa8b-41cd-babc-20e098597a11n%40googlegroups.com.
<error result latent choice model.jfif><error result latent choice model-1.jfif>

Bierlaire Michel

unread,
Nov 28, 2022, 2:17:34 AM11/28/22
to azis....@gmail.com, Bierlaire Michel, Biogeme
It looks like a specification issue. 
Did you check the eigenvector analysis at the end of the HTML file? It usually reveals overspecification issues. 


On 28 Nov 2022, at 04:39, azis....@gmail.com wrote:

yes, i run it using sequential at first... after that i used it for initial value for full estimation....
the result still show error 1.8e+308...

do you have idea for solving this problem?
--
Best Regards,

Azis Hakim Sjafruddin
Institut Teknologi Bandung

Reply all
Reply to author
Forward
0 new messages