Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: re: missing dummy variable


From   "Wallace, John" <John_Wallace@affymetrix.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: re: missing dummy variable
Date   Wed, 10 Oct 2007 10:59:28 -0700

Dear Statalisters,

I have been exploring -xi3- recently, and I've been wondering why the e.
option for generating indicator variables relative to the grand mean
doesn't work the way Kit and Maarten describe.  I've been playing with
some toy datasets trying to figure out how to generate the 95%
confidence intervals for the coefficient of the dropped level.  I've
managed to deduce how to calculate the coefficient itself, but the
standard errors and so forth aren't obvious.  Expanding on Maarten's
example (with Kit's suggestion included):

*------------------ begin example ---------------
sysuse auto, clear
gen domestic = !foreign
rename make desc
gen make = word( desc,1)
tab make

reg mpg foreign domestic, noconst hascons
xi3: regress mpg e.foreign
xi3: regress mpg e.make
*------------------ end example ---------------

Its clear that the information in the first two regressions is
equivalent, but the display would be better still if it was a
combination of both!  If the _cons from the second regression was
combined with the explicit listing of both indicators in the first.  

When the list of indicator variables gets long (as in the third
regression) , and especially when there are interactions,  the task of
interpreting the dropped variables (and especially significance in the
t-test relative to the grand mean) becomes non-trivial (to me at least).

Is there something I'm missing?

John W.


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Kit Baum
Sent: Thursday, October 04, 2007 6:20 PM
To: statalist@hsphsun2.harvard.edu
Subject: st: re: missing dummy variable

Maarten suggested

You may have noticed that in the second model (both dummies) I did not
have a constant in the equation, this is delibarate. If you want to
enter all dummies you have to exclude the constant. The way to do that
is to add the -noconst- option in Stata, see the example below:

*------------------ begin example ---------------
sysuse auto, clear
gen domestic = !foreign

reg mpg foreign
reg mpg foreign domestic, noconst
*------------------- end example ----------------

In the last command, I would use

reg mpg foreign domestic, nocons hascons


By adding the -hascons- option you are telling Stata that there  
really is a constant term in the equation by virtue of using the  
complete set of dummies. This will cause the F-stat of the second reg  
to be exactly like that of the first, as it is testing the hypothesis  
that the coeffs differ from the grand mean of mpg rather than from zero.



Kit Baum, Boston College Economics and DIW Berlin
http://ideas.repec.org/e/pba1.html
An Introduction to Modern Econometrics Using Stata:
http://www.stata-press.com/books/imeus.html


*
*   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/



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