Correction term for alternative sampling

[shad...@gmail.com]

Dear Michel,

I'm estimating an MNL model with over 1000 alternatives, so I need to perform sampling of alternatives for estimation. For each observation, I want to categorize the alternatives into 3 categories based on alternatives' distances from the origin observation, and sample from each category a different sample size. Therefore, the categorization is different for each observation.

Is there a function in Biogeme that performs this?

Currently, the example I found is when the partition is constant for all observations (like restaurant types).

If not, assume I have 1000 alternatives (including the chosen), and in category 1 I have 200 alternatives which I want to draw from there 10, and the chosen alternative belongs to this category, so in fact I need to draw 9, then the correction term for each of these sampled alternatives (and the chosen one) will be 10/200. Am I correct?

Thank you very much for your help,

Shadi