Re: st: multiple regression with dummy variables

[David Hoaglin <dchoaglin@gmail.com>]

Alex,

If you take advantage of Richard's suggestion, the resulting
coefficients will give you an indication of whether the effects of the
two sets of categories are additive.  In that connection, you can also
test whether the interaction effects as a group (4 degrees of freedom)
make a significant contribution.

Even if you do not need the interaction effects, you should interpret
the coefficients for each of those sets of three variables as the
effect of being in the particular category, adjusting for the
contribution of the other set of three variables.

David Hoaglin

On Wed, May 15, 2013 at 1:19 PM, Alexander <alex.lee.kotra@gmail.com> wrote:
> Dear Statalist readers,
>
> I am conducting a multiple OLS regression of some profitability
> measure of several companies on 2 sets of dummy variables.
>
> The first set consist of 3 variables which take 0 or 1 values
> depending on the size of a company: small, mid, large
> The first set consist of 3 variables which take 0 or 1 values
> depending on some other characteristic: small2, mid2, large2
>
> Each set encompasses all of my dataset (i.e. all of the companies are
> either small, mid or large, and all are either small2, mid2, large2)
>
> I understand that if I use all 3 variables of a set in the regression,
> one of them is omitted. In this case I could either get the omitted
> variable's coefficient out of the constant term or suppress the
> constant term.
>
> However, my purpose is to find out all the regressor's (out from both
> sets) coefficients in a single regression, that is:
>
> -regress PROFIT small mid large small2 mid2 large2
>
> As a result STATA omits two variables (small and small2), therefore I
> do not understand how to make sense of the results, cannot find the
> coefficients for the omitted variables and ultimately cannot make any
> inference with regards to connection between size, other
> characteristic and profitability measure.
>
> I would appreciate any help.
>
> Thank you.
> Alex
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/