Re: st: Thread-Index: AcqwrkqsyVJMieBERUiHkK2aIfQufQ==

[Austin Nichols <austinnichols@gmail.com>]

Jia Li <jli@arec.umd.edu>:
The new -gmm- command in Stata 11 is probably a better way to go,
assuming you have a lot of data.  You can specify a number of moment
conditions corresponding to the identifying assumptions of your
various equations, and you don't need to make such strong assumptions
about the error distributions.  But see also -cmp- on SSC.

On Thu, Feb 18, 2010 at 10:23 AM, Jia Li <jli@arec.umd.edu> wrote:
> Dear Statalisters:
>
>
>
> I would like to get some help/suggestions on a problem I am working on but couldn't solve.
>
>
>
> My research involves modeling household behavior with respect to short-run energy demand and long-run choice of energy technology. Both decisions flow from the same underlying preferences, thus the solutions to the short-run and long-run problems share the same parameters and the error terms are correlated. The short-run demand solutions (budget share) are nonlinear and the long-run discrete technology choice probabilities are derived from logistic distributions.
>
>
>
> I have separately modeled the short-run budget share equations using -nlsur- (assuming normal distribution) and the long-run model using mixed logit but need to tie the two components together. The estimation strategy includes 2SLS, 3SLS and FIML. However, I am not sure how to set it up in STATA, particularly considering I have both continuous and discrete dependent variables. It seems that I need to write my own likelihood function. But how should I do it when the assumed distributions are different?
>
>
>
> I'd appreciate any thoughts or suggestion you may have. In addition, I'd also be interested in examples that you think could help me.
>
>
>
> Thank you so much and I look forward to hearing from you!
>
>
>
> Sincerely,
>
>
>
> Jia

*
*   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/