Re: st: Interaction term with categorical variables and no. of observations stepwise regressions

[Maarten Buis <maartenlbuis@gmail.com>]

On Tue, Jul 31, 2012 at 11:39 AM, Marcel Steinter wrote:
> I have a panel data set (bond price - years)
> with 3 time periods of interest in my data (coded with 2 dummy variables). In
> addition to potentially including these 2 dummies I am also interested in
> interaction effects, in particular I would expect the influence of my main
> independent variable to be higher in the third period. I therefore include the
> interaction main independent variable x dummy for period 3 in my regression.
>
> Do I also need to include the interaction
> of the main independent variable x dummy for period 2 in my regressions (as for
> normal dummies when all have to be included). I would assume that I definitely
> have to include the main effect dummies 1+2.

That depends. If you leave interaction effect with the indicator
variable for t=2 out you are saying that the effect of your
independent variable is the same in t=1 and t=2, and differs from
these two in t=3. If you are happy with that description than you
should leave the interaction effect for the indicator variable t=2 out
of the model. If you want the effect of your independent variable to
differ in t=1, t=2 and t=3 than you should also include that
interaction effect.

Hope this helps,
Maarten

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany


http://www.maartenbuis.nl
--------------------------
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/