Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, statalist.org is already up and running.

# st: mlogit margins

 From Ida Hansen To statalist@hsphsun2.harvard.edu Subject st: mlogit margins Date Thu, 20 Sep 2012 13:58:20 +0200

```Dear Statalist,
I have estimated a multinomial logistic regression with 4 outcomes,
separately for immigrants and natives. My purpose is to examine if the
negative effect of dropping out on the outcome variable are more
negative for immigrants than for natives. I have computed AME, and in
absolute terms, there are no differences between immigrants and
natives (see analyse 1 below, shows the dofile for immigrants).

However, I have also computed predicted AME for dropouts (drop_out=1)
and for school completers (drop_out=0), (see analyse 2, shows the
dofile for immigrants). Comparing the relative differences between
school completers and school dropouts for native and immigrants, the
results are somewhat different. However, my first question is if the
command in analyse 2 is correct if I want to get predicted AME for the
two different values on the variable drop_out. My second question is
if it is correct to speak of predictive margins or is it another
proper name for the numbers you get in this table (analyse 2). My
final question is how I can compute a t-test to test if the relative
differences between dropouts and school completers are different in
the two samples (natives and immigrants).
I would appreciate any help!

Best regards,
Ida

Analyse 1.
mlogit outcome  kvinne drop_out alderdum2 if immigrant==1, base(0)
margins, dydx(*)predict(outcome(1))
margins, dydx(*)predict(outcome(2))
margins, dydx(*)predict(outcome(3))

Analyse 2.
margins, at(drop_out=1 drop_out=0)predict(outcome(0))
margins, at(drop_out=1 drop_out=0)predict(outcome(1))
margins, at(drop_out=1 drop_out=0)predict(outcome(2))
margins, at(drop_out=1 drop_out=0)predict(outcome(3))
*
*   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/
```