Dear statalisters,
further to the code provided by Sophia Rabe-Hesketh to Kristian Karlson as a solution to his problem for simulataneous Multinomial Logit estimation with two Multinomials, (SV: st: Multinomial transition model with unobserved heterogeneity) would anybody have any idea about estimating a gllamm multinomial model in expanded form for a binary (two categories) dv, using a three expanded options dataset?
The overall goal would be to estimate -what cannot be estimated according to gllamm manual- a mixed binomial - multinomial model, without using the fv() lv() specification which is not allowed for expanded datasets options in gllamm.
By using three categories of expanded data but a binary dependent variable for the first (alt=1) category with all the other (alt=2,3) observations equal to zero, the model estimates are not correct because computations are based on the full sample of three expanded categories and not of only one category -which could be e.g. obtained from a "gllamm l(logit)" specification for the one third of the observations.
As this code recently provided for two multinomials getting estimated simultaneously sounds promising, I wonder whether anyone had some suggestions on the problem I am encountering.
Many thanks,
Dino K.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
