Joint estimation of arrival rate and discrete choice parameter

[Philipp Theile]

Dear Biogeme-Group,

I am using revealed preference data of EV-charging choices. As I only observe positive choices, I cannot distinguish no-purchases from no-arrivals. For a similar case, Newman et al. (2014) propose a model, that fits a Poisson-distribution to the arrival rate to distinguish no-arrivals from no-pruchases (see the attached text snippet). So, the discrete choice model estimates parameters of the discrete choice, the no-purchase alternative, and the arrival-rate distribution.

I wonder whether I can implement such a likelihood function with Biogeme's tool set, or whether I would have to write the estimation from scratch. Do you have an intuition about these kinds of models?

Kind regards

Philipp

Newman, Jeffrey P., et al. "Estimation of choice-based models using sales data from a single firm." Manufacturing & Service Operations Management 16.2 (2014): 184-197.

[Michel Bierlaire]

You simply need to write the contribution to the loglikelihood function of each observation, which is Eq (5).

> --
> 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/751b341b-5c03-492f-a191-a8786fdab091n%40googlegroups.com.
> <Screenshot 2024-07-10 114023.png>

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