Luca - If you're doing weighted regression, check p.89 of the Q-St manual
under -regress-. By using the "hat" option with _predict, you can get h_j,
after which s.e.(r_j)=s*sqrt(1-h_j) where s is the root residual mean
square. The "hat" option appears to work for weighted regression. However, I
haven't thoroughly checked to see if the "hat" option really gives the
correct variance estimate of the j-th predicted value when there are
weights. HOwever, a quick comparison with unweighted and weighted regression
shows that "hat" values do indeed change when weights are used - so maybe
they're correct.
Al Feiveson
-----Original Message-----
From: Luca Mancini [mailto:[email protected]]
Sent: Friday, December 27, 2002 11:11 AM
To: [email protected]
Subject: st: residuals' se after weighted least squares
Hi,
I need to calculate the standard errors of the residuals (for each
observation) after weighted least squares. I noticed that the Stata
'predict stdr' option does not work in this case. Any thoughts on how
to get round this?
Thanks
luca
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Luca Mancini
Department of Economics
University of Warwick
Phone: +39 328 09 52 193
Webpage: http://www.warwick.ac.uk/~ecrfm/
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*
* 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/
*
* 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/
