multinomial grouped logit

[franck]

hello rats users,

I would like to know how estimate a multinomial gouped logit model in RATS with seven alternatives. We can't use the function "DDV" with grouped data (only with individual observations). Can I use the six odds ratio regressions (ln Pi/Pj = bX) by OLS or have you a procedure to do this type of model ?

Thanks a lot for your help