I ran the following logit model and then computed the marginal effects of my variable of interest in this way.
Code:
svy, subpop(if marital==1&age>17&age<35) : logit event1 ib0.skills##i.agesm i.sex i.educ2 i.wave5 i.biosonsn2 i.london ib1.nssecpar i.religion3 i.ethnicity, or level(95) margins r(0 1 2 )b0.skills, level(95) atmeans post at(agesm=(0(1)3)) subpop(if marital==1&age>17&age<35) noatleg
Now I would like to understand whether the differences that I obtained are signficantly different from one reference age group (18-21). I want to test it formally. I am using this code:
collect : margins r(1 2 3).skills#r.agesm, level(95) atmeans post subpop(if marital==1&age>17&age<36&(evunion==2)&jbstat!=7&iv fioall==1&save_new<2&skills<5&contr3<3&tercile3<3& means_benefits22<3&finnow3<2&finfut<4&tenure3<4&in du!=1) saving(skills, replace)
obtain this output
Code:
Delta-method
Contrast std. err. [95% conf. interval]
skills#agesm
(Routine vs Non-routine) (22–24 vs 18–21) -.0075323 .0154369 -.0377924 .0227279
(Routine vs Non-routine) (25–29 vs 18–21) -.0269329 .0165922 -.0594576 .0055919
(Routine vs Non-routine) (30–34 vs 18–21) .0189692 .0197626 -.0197703 .0577088
(Not employed vs Non-routine) (22–24 vs 18–21) -.0767736 .0162389 -.1086059 -.0449413
(Not employed vs Non-routine) (25–29 vs 18–21) -.0604068 .0207514 -.1010845 -.0197292
(Not employed vs Non-routine) (30–34 vs 18–21) -.0243907 .018812 -.0612668 .0124853
I interpret the fourth line saying that "The differences in the probability of Y between not employed and non-routine are signficantly greater among those aged 22-24 than among those aged 18-21". In sum, this means that the marginal effects of not employed vs. non-routine increase with age.
I would like to know if I interpret this in the correct way.
Thank you.
Lydia
0 Response to Interpretation r.a#r.b interaction (margins command)
Post a Comment