ICLV with mixed logit and panel effect

[Sally Chen]

Hi Michel,

I'm trying to estimate an ICLV, combing with panel effect and whose core model captures a nesting structure. I have a couple questions. 

I have no issue estimating it without panel effect, with a nested logit model as the choice model. However, I wasn't sure how to convert that to a mixed logit with error component and how to add the panel effect. 

1. I saw a post before in a similar situation. However in another post, I saw you mentioned that once the data is declared panel, "the random variables are distributed across individuals". How does Biogeme tell which one differs across individual, and which one differs across observations? 

2. In this example, the data has not been declared as panel data. Is the panel effect captured by having ec_sigma in both structural and choice model? 

3. If 2 is the correct way to capture panel effect, can I add another error component to the choice model to capture the nesting structure? Will that random variable be distributed across observations automatically? Or would you suggest using nested logit instead? 

Thank you so much for answering all these questions in advance!

Best,

Sally