RE: st: Blundell & Robin (1999) estimator

[ben jones <jonesblw@yahoo.com>]

Hi

I have estimated a Quaids demand system using GMM but the results are too slow to compile. Blundell and Robin (1999) and Lewbel and Pendakur (2009) both use a form iterated linear estimator which is supposedly quicker to run. 

This issue has been the subject to previous post, and the code employed by Lewbel and Pendakur (2009) has also previously been shared (see links below). However, I am currently struggling with a few basic issues:

First, in order to implement the Blundell and Robin estimator, should I iterate on FGLS rather than 3SLS as undertaken by Lewbel and Pendakur (2009)?

Second, if I wish to make weaker assumptions than independence between regressors and error terms, while using a control function approach to deal with exogeneity, what adjustments do I need to make to the standard errors?  

Third, I am slightly unclear how to code up the iterations on b(p) and a(p)? I cant see how the gamma matrix is determined in the Pendakur code, for example.

Any advice suggestions would be much appreciated.

Thanks!

Ben 




http://www.stata.com/statalist/archive/2011-10/msg00431.html

http://www.sfu.ca/~pendakur/iterated_3sls_without_pz,py,zy.do


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/