How to properly model and identify panel effect in Biogeme?

[张铎琪]

Dear all,

I’m currently working with panel data in Biogeme and trying to understand how to incorporate panel effects into my model. I have reviewed the example Mixture of logit with panel data, which is the most relevant reference I could find.

From the example, I see that the PanelLikelihoodTrajectory is used, and the documentation states that this enables working with panel data. However, I’m a bit confused about what exactly this function does. Is it just a way to correctly compute the likelihood for panel data, or does it also account for the panel effect (e.g. individual-specific unobserved heterogeneity across choices)?

Moreover, the results show sigma as a parameter from the mixed logit model, but I don’t see any explicit output or parameter capturing the panel effect per se. So I wonder:

1. Where is the panel effect reflected in the model results?

2. If I want to include individual-specific random intercepts (or other random parameters) to account for panel effects, how do I specify that in Biogeme?

Any clarification or example code would be highly appreciated!

Best regards,
Duoqi Zhang