Hi Statalisters. I've encountered what seems to me like inconsistent
behaviour of -regress- and -ereturn,post- in Stata (8.2 and 9.1) when a
variance-covariance matrix is not positive definite.
Specifically:
1. I estimate a model using -regress- with the -robust- option. There is
an obvious problem with the results: lots of coefficients have t-stats
that are virtually zero (but no standard errros are missing), and the
F-stat is missing. It turns out that the var-cov matrix is not
positive-definite.
Behaviour of -regress-: report output; no exit with error.
2. I estimate the same model with a user-written linear regression eclass
program. The coefficients and var-cov matrix are identical to case (1)
above. The program reaches the point where the coefficients and var-cov
matrix are posted with
ereturn post `b' `V'
Behaviour of -ereturn-: exits with error, namely
matrix not positive definite
r(506);
This seems inconsistent to me. Either -ereturn- should be able to post a
non-positive-definite var-cov matrix, or -regress- should exit with an
error and complain that the var-cov matrix is not positive definite.
Anybody else out there have views on this?
Cheers,
Mark
Prof. Mark Schaffer
Director, CERT
Department of Economics
School of Management & Languages
Heriot-Watt University, Edinburgh EH14 4AS
tel +44-131-451-3494 / fax +44-131-451-3294
email: [email protected]
web: http://www.sml.hw.ac.uk/ecomes
__________________________________________________________________
DISCLAIMER:
This e-mail message is subject to http://www.hw.ac.uk/disclaim.htm
__________________________________________________________________
*
* 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/
