Re: st: panel probit-tobit two stage model?

[Conan Li <conanli12@gmail.com>]

Hi Austin,

Thanks for the reply.

I have read the materials you provided. It seemed that the MSMM-gmm or
its equivalent "ivpois" is what I am looking for because I have some
IVs (z1-z5) to explain the adoption of y2, which had significant
values using "hetprob". But I am not sure if ivpois is applicable for
panel data like mine, which involves 50 states over 10 yrs of period.

Conan

On Tue, Jun 12, 2012 at 12:43 AM, Austin Nichols
<austinnichols@gmail.com> wrote:
> Conan Li <conanli12@gmail.com> :
> see also
> http://www.stata.com/meeting/boston10/boston10_nichols.pdf
> http://repec.org/usug2011/UK11_palmer_presentation.pdf (p 8, or 12 of 34)
>
> Do you have excluded instruments?
>
> On Mon, Jun 11, 2012 at 2:57 AM, Tirthankar Chakravarty
> <tirthankar.chakravarty@gmail.com> wrote:
>> Conan,
>>
>> A conditional likelihood RE estimator for the type of model you have
>> (see Section 3) is proposed in Vella and Verbeek (Journal of
>> Econometrics, Volume 90, Issue 2, June 1999):
>> http://dx.doi.org/10.1016/S0304-4076(98)00043-8
>>
>> A bias-corrected FE estimator for your model (see Example 2) has been
>> proposed in Fernandez-Val and Vella (Journal of Econometrics, Volume
>> 163, Issue 2, August 2011):
>> http://dx.doi.org/10.1016/j.jeconom.2011.03.002
>>
>> Note that the second estimator works with large T, so it might not be
>> suitable for your case. To my knowledge these are not available as
>> Stata packages, and you might need to code these up yourself.
>>
>> T
>>
>> On Sun, Jun 10, 2012 at 9:52 PM, Conan Li <conanli12@gmail.com> wrote:
>>> Hi Statalist,
>>>
>>> I am developing a model to estimate a particular state policy on state
>>> wind installation over 10 years of period. The tentative model for the
>>> wind installation is a tobit as: y1(i,t)=a+by2(i,t)+cx(i,t)+u, where
>>> y1 is the wind installation measured in MW, y2 is a dummy indicating
>>> the existence of the policy and x is a group of control variables. The
>>> policy variable is believed as endogenous because its adoption depends
>>> on some other variables. In this framework, how can I develop a model
>>> that integrates a first-stage probit to estimate the y2, and a
>>> second-stage tobit to estimate the y1?
>>>
>>> Here are some of my thoughts for the model:
>>> 1. ivtobit? (seemed not applicable if y2 is not continuous)
>>> 2. xtprobit->xttobit (how do I get there?)
>>>
>>> Thanks for your help!
> *
> *   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/



-- 
Qinghua Li (Conan)
State University of New York-College of Environmental Science and Forestry
MS Candidate in Environmental Studies
qli16@syr.edu

*
*   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/