Thank you very much for you suggestions, Arne and
Stas.
The dependent variable in the first stage is a proxy
for ownership rights to land. In this case, 0
corresponds to no ownership rights (only user/renter
rights), 1 to partial ownership rights (with
stipulations on sale), and 2 to full ownership rights.
I have an instrument (z) to solve the endogeniety of
the ownership rights variable in the second stage
regression.
If there is another (easier??) approach that would fit
this application, I would be very grateful to hear it.
Otherwise, I will look into how to use Mata.
Thanks again!
Xinyi
--- Arne Risa Hole <[email protected]> wrote:
> Hi Xinyi,
>
> I think you would be better off using Mata in this
> case, as Stas
> suggests. What makes things more complicated in your
> model compared to
> the examples in my paper is that the predicted
> values from the first
> stage model depend on the auxiliary parameters/
> cut-off points along
> with the index function which makes it difficult to
> express the C
> matrix in -matrix accum- terms. It may be possible
> to do it but
> calculating the scores for each variable in Mata
> seems to me to be an
> easier solution in this case. Mata's
> element-by-element multiplication
> operator :* will be helpful.
>
> I agree with Stas that the ordered probit in the
> first stage seems a
> little odd but of course I don't know anything about
> your application.
>
> Good luck!
>
> Arne
>
> On 17/01/07, Stas Kolenikov <[email protected]>
> wrote:
> > Check Hardin's article first, he gives a general
> treatment and shows
> > why the trick with -mat accum- is usable. You
> would have to take the
> > derivatives with respect to the first stage
> parameters properly, and
> > don't forget about the -oprobit- thresholds, too.
> If you are lucky,
> > you will just have a more complicated expression
> for the [iweight=]
> > with both zz1 and zz2 that would be offset by the
> difference in
> > thresholds somehow, but it may happen that what
> you'd get won't be
> > expressable in -mat accum- terms, and you would
> have to program that
> > in Mata using -st_view()-s (which tends to be more
> elegant anyways).
> >
> > I personally don't think the first stage -oprobit-
> is a very good
> > idea... but I am not your advisor/reviewer, anyway
> :)).
> >
> > On 1/17/07, X W <[email protected]> wrote:
> > > I am having some trouble figuring out how to
> extend
> > > the Murphy-Topel correction for standard errors
> in a
> > > two-step model with an ordered probit in the
> first
> > > stage and a logit in the second stage.
> >
> > --
> > Stas Kolenikov
> > http://stas.kolenikov.name
> > *
> > * For searches and help try:
> > *
> http://www.stata.com/support/faqs/res/findit.html
> > * http://www.stata.com/support/statalist/faq
> > * http://www.ats.ucla.edu/stat/stata/
> >
> *
> * For searches and help try:
> *
> http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
____________________________________________________________________________________
Do you Yahoo!?
Everyone is raving about the all-new Yahoo! Mail beta.
http://new.mail.yahoo.com
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
