Kathleen,
The use of a binary endogenous variable could pose problems in a simultaneous
equation model, beyond the sem bias corrected for by efficient 3sls. You have
range restriction that would attenuate the relationship. The residuals would be
quite nonnormal. If the proportions in the two codes were skewed, this effect
could be aggravated. Would 3SLS adequate correct for this? Perhaps somewhat.
OLS on such a variable tends only to work within certain limits, which would
not apply to more continuous endogenous variables.
You might consider constructing a polyserial-polychoric correlation matrix as input
to path model (or a confirmatory factor analysis with a path analysis between factors)
as an alternative. This might be possible with GLLAMM and LISREL.
Regards,
Robert
Robert A. Yaffee, Ph.D.
Research Professor
Silver School of Social Work
New York University
Biosketch: http://homepages.nyu.edu/~ray1/Biosketch2009.pdf
CV: http://homepages.nyu.edu/~ray1/vita.pdf
----- Original Message -----
From: "McCullough, Kathleen" <[email protected]>
Date: Monday, April 13, 2009 7:22 pm
Subject: st: Endogenous variables in a system
To: [email protected]
> Hi All, I appreciate your help in advance. Is there a way to either
> specify that one of the equations in reg3 has a binary dependent
> variable that is endogenous in the other equations or to alter cdsimeq
> to include more equations with continuous variables?
>
> Thanks!
>
> *
> * 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/
