st: Fwd: Error standards for margins effects in multinomial logistic regression

[Beatriz Rodriguez Prado <bprado@eco.uva.es>]

Hello,
I am running a multinomial logistic regression and in order to interpret
the results I have calculated the average margin effects. All variables
are treated as factors and there are also interactions.
More precisely, this is the code.


    #delimit;
    mlogit grupos experiencia_g2##sexoSS experiencia_g2##edad_SSr
experiencia_g2##menor2_r experiencia_g2##tip_conr
experiencia_g2##grup_co_r experiencia_g2##act_eco_rf
experiencia_g2##trab_ccs_r experiencia_g2##nivel_est1_r if grupos<5,
vce(robust);

    #delimit;
     margins, dydx(*) at(experiencia_g2=(1 (1) 4)) predict(outcome(1))
vsquish vce(unconditional);
     margins, dydx(*) at(experiencia_g2=(1 (1) 4)) predict(outcome(2))
vsquish vce(unconditional);
     margins, dydx(*) at(experiencia_g2=(1 (1) 4)) predict(outcome(3))
vsquish vce(unconditional);
     margins, dydx(*) at(experiencia_g2=(1 (1) 4)) predict(outcome(4))
vsquish vce(unconditional);


In the logistic regression some coefficients appear as not significant,
whereas the marginal effects are significant. Why is that possible? In
order to analyse the results, which ones sholud I consider?

Thank you very much
Beatriz Rodriguez




*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/