RUM vs RRM in biogeme: Is it possible Final LL, AIC, BIC, and confusion matrix are exactly the same?

[Natchaphon Leungbootnak]

Hello,

I have run 2 models: Random Utility Maximization (RUM) logit and Random Regret Minimization (RRM) to predict and understand the traveler choices between ML and GPL. 

The results of both models surprisingly showed me that Final LL, AIC, BIC, and the confusion matrix are exactly the same. The difference I observed, RRM had around 3 times longer optimization time than RUM.

In running both analyses, I use the same, including train and test datasets and logit models. The difference is that only RUM uses the utility function, and RRM uses the regret function (see in attachments).

I wonder why both analyses provided me the same Final LL, AIC, BIC, and confusion matrix? Should they be a little different? Could you please guide me about this?

I have attached code and result files for both RUM and RRM.

Best regards,

Natchaphon Leungbootnak

[Michel Bierlaire]

You've made a mistake. You have estimated twice the exact same model.
Also, the concept of "confusion matrix" does not make any sense for discrete choice.

> --
> 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 visit https://groups.google.com/d/msgid/biogeme/25645fbf-53f1-4d79-aa77-15e0dd0591bdn%40googlegroups.com.
> <RRMbiogeme.py><RUMbiogeme.txt><RUMbiogeme.py><RRMbiogeme.txt>