Re: AW: st: reporting results from a multiple imputed data set

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Wed, 20/4/11, Lamla, Bettina wrote:
> (1) i can only include  M-1 year dummies.
> If my command includes M year dummies anyway, stata will
> omit one dummy due to collinearity, i.e. the least
> significant one.
> (2) the least significant one varies across imputations.
> Hence I could just exclude let's say year 1 for all
> estimations and hence cannot report an estimate for this
> year
>  OR
> I let stata pick the year dummy to be excluded and as a
> result -after applying the rubin rule- have estimates for
> all M year dummies even though this would not be possible in
> the individual regressions

When you or Stata leaves out a dummy/indicator variable than
the remaining dummy/indicator variables represent comparisons
with that reference category. It matters whether you compare
a year with say 1991 or 2011: the coefficients will have a 
different interpretation. So if you try to combine combine
estimates that are based on different reference categories 
than that is a very bad thing: the results will be completely 
meaningless. So you must specify the reference category 
yourself, i.e. leave out one of the dummies.

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/