Re: st: calculation of marginal effect

[Maarten Buis <maartenlbuis@gmail.com>]

--- On Tue, Aug 28, 2012 at 8:39 PM, Yanru Qiao wrote me privately:
> We analysis the  the racial disparities in health care area, Like,
> blacks(whites) who are not eligible for a health care program are more
> likely to be sick than those who are eligible under different health
> care utilization measurements . And here is an example of my
> code(bnh--black, main06--eligible for program):
>
> svy: logit goodhealth i.bnh##i.main06, or nolog
>
> margins, over (bnh main06) expression (exp(xb()))  post
>
> lincom 0.bnh#0.main06-1.bnh#0.main06
>
> lincom 0.bnh#1.main06-1.bnh#1.main06
>
> lincom (0.bnh#0.main06-1.bnh#0.main06)-(0.bnh#1.main06-1.bnh#1.main06)
>
>
> My question is:
> 1.Can I use expression (exp(xb()))  to all other models? Like
> [R]regress [R]mlogit [R]nbreg?
>
> 2.How to interpret the marginal effect in a multinomial logistic
> regression (Like: we have five categories)

Such questions need to be sent to Statalist instead of individual
members. The reasons for that are discussed in
<http://www.stata.com/support/faqs/resources/statalist-faq/#private>

The code you quote comes from M.L. Buis (2010) "Stata tip 87:
Interpretation of interactions in non-linear models", The Stata
Journal, 10(2), pp. 305-308. The real point of that Stata tip is that
you should _not_ use marginal effects to interpret interactions but
odds ratios instead. So that is going to be the meta-answer to your
questions: If you have interactions, just forget about marginal
effects and learn to interpret your model in terms of the natural
metric of that model.

Ad. 1: No

Ad. 2: <http://www.stata.com/statalist/archive/2010-10/msg00149.html>

-- Maarten

---------------------------------
Maarten L. Buis
WZB
Reichpietschufer 50
10785 Berlin
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/