Re: st: ivprobit error message, outreg, lnsigma and athrho

["Brian P. Poi" <bpoi@stata.com>]

On Thu, 30 Nov 2006, Nirina F. wrote:

> Hello List,
> 1- I am running ivreg and it works fine but I would
> like to run ivprobit (as I have binary choice for my
> dependent variable) and it gives me the following
> error: inital vector: matrix must be dimension 126
> r(503). I don't know what to do to correct it.
Are you using the "from()" option to specify initial values?  If so, then 
you need to specify a row vector with 126 elements.

If you are not using the "from()" option, then this could be a bug with 
-ivprobit-.  In that case, if you send directly to me (bpoi@stata.com) 
your dataset along with the exact syntax you are using, I'd be happy to 
look into it.


> 3-Can somebody explain me the interpretation of the
> lnsigma and athrho please?
> Thanks a lot in advance,
> Nirina
lnsigma and athrho are transformations of sigma and rho that are used in 
the estimation process.  What you care about are sigma and rho, reported 
below lnsigma and athrho in the output.  Sigma is the variance of the 
reduced-form equation for the endogenous regressor.  The more interesting 
parameter is rho, which is the correlation between the errors in the 
probit equation and the reduced-form equation for the endogenous 
regressor.  Rho being equal to zero is equivalent to saying that the 
variable you suspect to be endogenous is in fact exogenous; in other 
words, if the estimated rho is insignificant, then we cannot reject the 
null that there is no endogeneity issue and a plain probit regression 
could be used.

  -- Brian Poi
  -- bpoi@stata.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/