Re: st: mfx for stata 12

[Richard Williams <richardwilliams.ndu@gmail.com>]

At 02:50 AM 8/24/2012, Chiara Mussida wrote:
> Dear all,
> i estimated a mlogit with 8 outcomes, which reject the IIA assumtion
> tested with a hsiao test:
> mlogtest, sm
> i would be pleased to get the marginal effects of my model. If i run
> the command to obtain, for instance, the marginal effects for outcome
> 1:
> mfx compute, predict(outcome(1))
> i obtain very suspects results, namely mfx with signs and sign
> completely different from the ones of the  coefs estimates.
> I use Stata 12, and from the manuals mfx might compute the marginal
> effect like in the previous stata versions. Do you know if something
> changed?
> Thanks

A few side points first:

* -mlogtest- is part of Long & Freese's -spost9- package (or at least 
it should be, as that is the most current version of the program; 
older versions are also floating around). The file is on Long's own 
site and can be found with -findit-.
* If you are going to use -mfx- rather than the newer -margins- 
command, you should probably use the user-written routines -margeff- 
or -mfx2-. They are easier to use and/or quicker after multiple 
outcome commands like -mlogit-. Both are available from SSC.

Now, getting to your question -- such sign flips have to happen with 
multiple-outcome commands like mlogit. If, say, increases in X2 make 
it more likely to be in category 2 than in category 1, then increases 
in X2 make it less likely to be in category 1 than category 2. That 
is, as you look at all outcomes (not just the first) you are going to 
see changes in sign and coefficients. To illustrate, try this after 
installing mfx2:

use "http://www.indiana.edu/~jslsoc/stata/spex_data/ordwarm2.dta";, clear
mlogit warm yr89 male white age ed prst, base(1)
mfx2, nolog

Finally, assuming you are using Stata 11 or higher, consider using 
the new -margins- command. It has many advantages (e.g. you can use 
factor variables, it will handle interactions and squared terms more 
easily). Its main drawback is that, as far as I know, there are no 
commands like -margeff- or -mfx2- that make it easier to handle 
multiple outcome commands. If I ever got wildly ambitious, I might 
try writing a -margins2- command, but I keep hoping StataCorp will 
write it for me.


-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
WWW:    http://www.nd.edu/~rwilliam

*
*   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/