--- ucb_gal <[email protected]> wrote:
> i'm still trying to figure out postestimation commands with mim and
> mlogit. i would like to compare coefficients between models (e.g.,
> how does the coefficent for income change in specifications with and
> without education?).
The reason why people do this is to see how much of the effect of
income can be explained by education. In the model without education
income will capture a part of the effect of education because education
and income are correlated. It is this part of the effect of income that
is being explained by education.
The problem is that this is not what you get with any non-linear model
((almost) any model except -regress-). If education has an effect on
your dependent variable than adding it will cause a change in the
effect of income (most likely to go up), even if education and income
are uncorrelated. This is most clearly explained in this course note by
Richard Williams:
http://www.nd.edu/~rwilliam/xsoc73994/l06.pdf. Problem is, he proposes
standardized coefficients as a solution, but that is not
straightforward to define in -mlogit-. So my advise is don't do this if
you do not need this to answer your main research question (people
often do this out of habit). If you can slightly revise your research
question so you won't need this, seriously consider that. If you really
need to do this, be prepared for a lot of studying and thinking.
> i would also like to compare the fit of two or more models (e.g. how
> much more variation is explained when i include new independent
> vars?).
In non-linear models this is typically done with a likelihood ratio
test. I gave an example on how to do that after -mim- in this post:
http://www.stata.com/statalist/archive/2007-04/msg00900.html. However,
I think I made an error in that code since it gives quite different
results depending on the seed (I ran it twice and it gave different
answers, than I looked at changing the seed and keeping the seed
constant, and it was the seed that caused these differences). It is a
lot of code, though pretty straightforwardly transcribed from (Schafer
1997) page 117. I can't check the code today, since I don't have the
book at home (today is a national holiday in the Netherlands,
celebrating our monarchy). I probably should be wise and not check it
tomorrow either, but instead work on my dissertation. So if you or
anyone else wants to have a look at it...
Schafer, J.L. (1997) Analysis of Incomplete multivariate data. Boca
Raton: Chapman & Hall/CRC
Hope I haven't discouraged you too much,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
___________________________________________________________
Yahoo! Mail is the world's favourite email. Don't settle for less, sign up for
your free account today http://uk.rd.yahoo.com/evt=44106/*http://uk.docs.yahoo.com/mail/winter07.html
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
