choice-metrics.com

Dear Michiel,

Nowadays, the use of hybrid choice models (incorporating latent variables into the discrete choice models) have been getting popularity in different research area.

Is there any experimental design syntax as an example coded by Ngene for a hybrid choice model (discrete choice model with latent variables)?

Best regards,
Nima

It is not possible to generate an experimental design for a hybrid choice model for several reasons:

1. Optimising a design for the panel mixed logit model is generally not computationally feasible (it may take months of computation time). Therefore, optimising for an even more complex model such as the hybrid choice model is simply not possible (not even on a super computer).

2. Obtaining parameters priors for the hybrid choice model would be nearly impossible.

The current state of the art is to optimise a design for the multinomial logit model using Bayesian priors. Such a design is generally also fairly efficient for more advanced choice models. I am not aware of any software that can optimise a design for the hybrid choice model.

Michiel