Log-likelihood magnitud with panel data

[Adrian Molescu]

Hello,

I am using the Parking SP data from Spain, available in the Biogeme website, to estimate some models.

I have a simple model with a random parameter, and I'm trying to add panel effect (the data has 8 responses for 197 individual, so 1756 observations in total). I am using the expression PanelLikelihoodTrajectory and the model is estimating without issues.

The model with a random parameter without panel effect, and all other base models (MNL, CNL, ML, etc), have Log-likelihoods of around -1,090 to -1,110

The model with a random parameter and panel effect has a Log-likelihood of around -487

How can I interpret such a change in the objective function? How is the Log-likelihood computed when panel effect is included?

I've checked the documentation, and for what I understood the model with PanelLikelihoodTrajectory should also be considering the 1756 total probabilities, just like the others.

Many thanks in advanced,

Adrian

[Bierlaire Michel]

If the data is declared panel, one observation corresponds to the trajectory one individual. If the data is not declared panel, each row corresponds to one observation.

> --
> 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/ccc15be6-4044-4480-8e9b-5013e1683b21n%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