Re: st: Nlsur quaids and IV

[Jorge Eduardo Pérez Pérez <perez.jorge@ur.edu.co>]

To do this, you would need to have a estimator for nonlinear 3SLS
which Stata does not have. If you want to keep using Stata, there are
some additional approaches you might try though:

* Estimate the model via an interative estimator and -reg-, like in
Blundell and Robin (1999). You would have to include the residuals
from the first stage regression in the share equations to correct the
endogeneity (This is done in Banks, Blundell & Lewble QAIDS paper).
Blundell and Robin provide a formula for the standard errors, but
since the estimator is computationally efficient, it is easier to
bootstrap the whole thing.
* Estimate the model with -nlsur- including the residuals from the
first stage regression in the share equations. Again, adjustment of
the standard errors is needed. I am not aware of the adjustment in
this case. Bootstrapping would be slower.
* You can estimate a -LA AIDS- with full correction for endogeneity
and correct standard errors using -reg3-, no iterations would be
needed.
* You can estimate the AIDS or QAIDS with -reg3- and an iterative
estimator like the one mentioned before. Some adjustment of the
standard errors is needed for estimation with generated regressors.
* You can estimate the model via -gmm-


References:

Blundell, Richard & Robin, Jean Marc, 1999. "Estimation in Large and
Disaggregated Demand Systems: An Estimator for Conditionally Linear
Systems," Journal of Applied Econometrics, John Wiley & Sons, Ltd.,
vol. 14(3), pages 209-32, May-June


_______________________
Jorge Eduardo Pérez Pérez




On Fri, May 27, 2011 at 7:18 PM, Ana <analunhb@gmail.com> wrote:
> Dear Statalisters (Brian Poi and Alex Olssen)
>
> I'm using Brian Poi (2008) nlsurquaids.ado to estimate a Four Demand
> Equation system.
>
> My model has the form:
>
> nlsurquaids @ w1 w2 w3 lnp1-lnp4 lnexp, ifgnls nequations(3) param(a1
> a2 a3 b1 b2 b3 g11 g12 g13 g22 g23 g33 l1 l2 l3) nolog
>
> I'd like to include in this model an Instrumental Variable (like
> lnincome) for the expenditure variable (lnexp). Is it possible?
> Does anybody know how to include the IV in this model?
>
> Thanks a lot in advance.
> Best regards,
> Ana.
> *
> *   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/
>

*
*   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/