The numerical derivatives are approximate flat or discontinuous region encountered in GMNL.

[Sukunta Muadthong]

Dear Statalist,
I want to check if my data has scale heterogeneity via a generalized multinomial logit model (G-MNL) in the discrete choice experiment model described by Gu, Y., Hole, A. R., & Knox, S. (2013). My code is

gmnl Y , rand( asc fish_d2 fish_d3 vet_d2 vet_d3 alc_d2 alc_d3 smk_d2 smk_d3 pqt_d2 pqt_d3) group(ncs) id (ID) nrep(500) scale(scale) gamma(0)

The program could not identify the model. It shows "numerical derivatives are approximate
flat or discontinuous region encountered" and backs up the iterations. Please, recommend to me how I do the next step.
Sincerely yours,
Sukunta

[Carlo Lazzaro]

Sukunta:
the usual recipe is to sart it all over again, adding one predictor at a time and see when Stata starts to complain.

[Sukunta Muadthong]

Dear Carlo Lazzaro,
Thank you so much for your reply. The discrete choice experiment can not be deleted the predictor (The x variable) from the model. However, I run the model with the syntax

gmnl Y fish_d2 vet_d2 smk_d2 pqt_d2 , rand( asc fish_d3 vet_d3 alc_d2 alc_d3 smk_d3 pqt_d3) group(ncs) id(ID) nrep(500) scale(scale) gamma(0)

It can estimate the model. What does it mean?
Sincerely yours,
Sukunta