ApolloChoiceModelling forum

Hi,

I'm wondering if there is any way to specify multi-level MNL? I guess it is related to the mixed logit model, but I'm not sure how to specify it.
For example, individuals live in different areas, and individuals living in the same area have the same random effect influencing the utility functions.
For individuals living in the first area, the utility function would be:
v[['car']] = (b_cost_car + random_carcost1) * cost_car + random_area1
v[['train']] = (b_cost_train+ random_traincost1) * cost_train + random_area1

For individuals living in the second area, the utility function would be:
v[['car']] = (b_cost_car + random_cost2) * cost_car + random_area2
v[['train']] = (b_cost_train+ random_traincost2) * cost_train + random_area2

Thank you!

Hi

it's not quite clear what you mean by multi-level in this case.

And in the below specification, the effect would drop out as only differences in utility matter.

Can you be more specific what you wish to achieve? Heteroskedasticity across groups?

Stephane

Hi,
Thank you for your reply.
Yes, the effect would drop out, as you said. Only one utility function would have the additional effect term in it. By multi-level, I mean there will be random effects caused by the group the individuals belong to, which cause additional random terms added in the utility functions.
Thanks!

How many such groups do you have?

Hi,
I have three such groups.

So then you could just estimate scale differences between the 3 groups. Like in the RP-SP example

I see, thank you!