Re: st: fmlogit interpreting marginal effects

[Maarten Buis <maartenlbuis@gmail.com>]

On Fri, Mar 16, 2012 at 6:23 AM, Trudy Sullivan wrote:
> I have six dependent variables that are fractions and add up to 1. I used fmlogit to estimate a fractional multinomial logit and I used dfmlogit to compute the marginal effects.

-fmlogit- and -dfmlogit- are user written programs, so per the
Statalist FAQ you must say where you got it from. This is not to be
mean, but to help you ask answerable questions. There are often
multiple versions of user written software floating around in cyber
space, and you can imagine that it helps a lot if everybody is talking
about the same piece of software. I am assuming you are using the
version downloaded from SSC. There are many other tips on how to ask
answerable questions/avoid common mistakes in the FAQ. A link to which
you can find at the bottom of this post (and every other post on
Statalist).

> In my case the six dependent variables could equally be used as the reference category. I want to interpret the marginal effects for the six dependent variables but depending on which variable is omitted, the marginal effects change slightly. How should I proceed?

The computations get quite involved, so I guess this is a precision
issue. So you can just pick one. There may be ways of changing the
computations to get more accurate results, but since I don't think it
is a big problem it will be low on my list of priorities.

> dmflogit computes the marginal effects for the omitted category and I don’t understand how to interpret this marginal effect.

You interpret them the same way as all the other marginal effects.
There are no separate coefficients for that proportion, since the
remaining coefficients fully determine how that proportion changes.
This is what is being used to compute those marginal effects.

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/