On Feb 19, 2004, at 2:33 AM, M. wrote:
Dear Friends,
Now i am doing a Hausman specification test, testing the difference
between GMM-instrumental variable estimates and OLS estimates for the
possible
endogenity of the regressors. In some cases i getting negative
Chi-square
value. (This is because the variance-covariance matrix of the
differeces of the
coefficients is not positive definite) How can I interpret it? In the
case of
negative chi-square values, can i accept the null hypothesis? Any
suggestions
andcomments are most welcome.
Thanking you in advance,
M Parameswaran
Suggest you read the Stata Journal (v3:1) paper by Baum, Schaffer,
Stillman on IV and GMM; it is also available from our dept webpage as a
working paper (http://www.bc.edu/economics). Our 'ivreg2' (which does
IV-GMM) performs an equivalent test which does not run afoul of the
problems that the Hausman test encounters in small samples.
Kit
*
* 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/
