Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: DFBETA for constant


From   Roger Harbord <Roger.Harbord@bristol.ac.uk>
To   statalist@hsphsun2.harvard.edu
Subject   st: DFBETA for constant
Date   Mon, 5 Aug 2002 20:50:28 +0100

Why can't I generate the DFBETAs for the constant after a regression? 
Neither -dfbeta- nor -predict, dfbeta()- will allow me to do this. 
-regress- labels the constant as "_cons", but obviously I can't do
. dfbeta _cons
or
. predict dfcons, dfbeta(_cons) 
after -regress-. I don't really see why a "DF_cons" isn't generated by 
-dfbeta- along with all the other DFs when no varlist is given.  After 
all, the constant is just as valid a regression coefficient as the 
others.

If I instead try 
. gen one=1
and include "one" along with my other regressors as an explicit 
constant, and add to -regress- the option:
1) -hascons-, then -dfbeta- generates DFone containing all missing 
values;
2) -noconstant-, then -dfbeta- reports the error "not appropriate after 
regress, nocons".  I'm afraid I don't understand why it's not 
appropriate.

As I understand it, DFBETA[i] is the change in the estimate of the 
coefficient when the ith observation is deleted, scaled by the 
estimated standard error of the coefficient.  I can't see where that 
would break down when the coefficient in question is the constant.

I appreciate it may be a little unusual to be interested in whether any
observations have a big influence on the estimate of the constant, 
but I really do want to know! 

I'd be very grateful for any explanations,
Roger.

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index