A noticeable difference in ASC.

[Amin Afshari]

Dear Prof. Bierlaire,

I am currently working on a hybrid choice model estimation and have encountered a notable difference in the alternative specific constants (ASCs) between the full model and a simpler model that only includes ASCs in the utility functions (representing the proportions). Specifically, in the simpler model, the ASC for the Auto mode is 1.39, whereas in the full model, the ASC for the same mode is 49!

I have checked the model specification and results and found no errors or issues. Should I be concerned about this significant difference, or is this an accepted outcome?

Thank you for your time and assistance.

Best regards,

Mohammad Amin Afshari Moez

[Michel Bierlaire]

ASCs are the mean of the error terms. Therefore, it is not unusual that they are sensitive to the specification.

However, it is a good reaction to find this suspicious.
Among the possible causes, I can think of an insufficient number of draws in Monte-Carlo integration or an identification problem.

> --
> 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/0e2b51cd-66e7-4b8a-96fc-ac394884366an%40googlegroups.com.

Michel Bierlaire
Transport and Mobility Laboratory
School of Architecture, Civil and Environmental Engineering
EPFL - Ecole Polytechnique Fédérale de Lausanne
http://transp-or.epfl.ch
http://people.epfl.ch/michel.bierlaire