Re: st: multiple imputation and propensity score

[Maarten Buis <maartenlbuis@gmail.com>]

On Wed, Aug 24, 2011 at 6:39 PM, Stefano Di Bartolomeo wrote:
> In truth I am trying to be humble and apply the best methodology I can.

The aim of a statistical method is to link the stuff you have seen
(the data) to a conclusion. It is thus part of a logical argument that
you are making. The best methodology is thus not the most complicated
or the method that controls for most "problems", but the one that
provides the clearest argument. In time your (sub-(sub-))discipline
will collect a great number of such arguments, and someone will digest
them to a "standard result". For individual studies it is more
important to do "their thing" right, the controlling for all possible
problems will happen at a later stage.

> As a last resort: do you think that using just one of the imputed datasets for matching + further analyses would be grossly inappropriate?

With that solution you would get wrong (too small) standard errors. It
would be better not to do imputation and just focus on getting your
matching model right.

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/