st: Re: dummy question

[Kit Baum <baum@bc.edu>]

An approximate solution to this question (as a little algebra will 
show) is to define the dummies 1-(G-1) as d_i - d_G for G groups. 
(Approximate depending on how group sizes differ across the dummies). 
As an example:

webuse grunfeld, clear
drop if company>5
tab company, gen(co)
forv i=1/4 {
  gen Co`i' = co`i'-co5

su invest
reg invest co1-co4
reg invest Co*
lincom -1*(Co1+Co2+Co3+Co4)

As you can see, the constant term in the latter regression is the grand 
mean.
The coefficient on the excluded dummy can be expressed (via lincom) as 
minus the sum of the other dummy coefficients.

Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html

On Nov 29, 2004, at 2:33 AM, Javier wrote:


> One possibility recommended in the literature was to
> use restricted least
> squres method (Greene, Seaks 1991). It recommends to
> set the restrictions to
> dummies so that the weighted (according to weights of
> observation in
> endogenous variable) sum of all dummies (no reference
> category is omitted) is equal to zero.
*
*   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/