Re: st: Interpreting negative Rho with bivariate probit (partial)

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Fri, 25/2/11, Jean-Christophe Boucher wrote:
> I am using a bivariate probit with partial observability to
> take into account the effect of a selection equation when
> the dependent variables are the same y1=y2. 
> 
> I have a hard time (and many discussions) in interpreting
> my Rho result which is negative and highly significant.
> Mostly, the negative direction is counter intuitive in my
> model, but some say that rho is only an estimation of all
> the unmeasured crap between both equations of the biprobit
> and I shouldn't be worried about it as long as it is
> significant.

rho is just the correlation between the residuals of the two
equations. So apperently there is a variable that is not in
your model that makes people have a success on y1 and at the
same time a failure on y2. I would not over-interpret this 
result though: if you were really interested in that variable
then you should have found a way to measure it and include it
in your model.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   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/