Re: st: behavior of -areg-

[Richard Williams <Richard.A.Williams.5@ND.edu>]

At 05:17 PM 2/27/2006, Radu Ban wrote:
> Dear Listers,
>
> I've encountered this puzzling behavior of -areg-:
>
> egen total_hh_nr = sum(hh_nr), by(keyid)
> ...
>
> areg y x1 x2 total_hh_nr, absorb(keyid)
>
> [by mistake I included total_hh_nr in the regression]. The puzzle is
> that when I run the regression "total_hh_nr" doesn't drop out (as I
> would expect, due to invariance within "keyid"). The standard error on
> "total_hh_nr" comes out very large, of the order 10^8, so there is
> something weird going on, but I cannot figure out what. Furthermore,
> the coefficient on "total_hh_nr" changes at every run of the program.
> So, what is going on in here?
I've found that things that should drop out, because of perfect 
collinearity, sometimes stay in, producing huge standard errors, 
implausible estimates, parameter estimates correlated .999, etc. Or, 
the program may iterate forever and not reach a solution.  I think it 
is because you do not have perfect precision in the numbers and hence 
you don't get perfect collinearity but you do get very close to it.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX:    (574)288-4373
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW (personal):    http://www.nd.edu/~rwilliam
WWW (department):    http://www.nd.edu/~soc  
*
*   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/