--- On Mon, 12/10/09, Dirk Deichmann wrote:
> Just for clarification, if I do what Maarten suggested I
> know whether a model improved in fit when the previously
> added variable turns out to be significant in the Wald test
> that I conduct after estimating the logistic regression.
If that test is significant then the larger model fits better.
This does not necesarily mean that that variable should be
included, think for example about the distinction between
confounding and intervening variables.
> But I do not know how much the model fit improved, right?
Than you would have to define what model fit means. There
are many possible ways in which you could define that, and
all off them are flawed (some more than others). Moreover,
a low model fit (however defined) does not mean a bad model,
it may just as well represent the state of the world: there
is lot that is unknown and possibly unknowable. Finally, you
always run the risk of doing evil things like data snooping,
stepwise, datamining, etc. For all these reasons I don't
think it is a very helful concept.
-- Maarten
--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany
http://www.maartenbuis.nl
--------------------------
Send instant messages to your online friends http://uk.messenger.yahoo.com
*
* 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/
