Combomarginsplot Logit Model

[Sean Vina]

I'm having difficulty using the combomarginsplot and I am not sure if it is because I am using a binary logit model. Below is the sample code followed by the individual outputs and the combomarginsplot.

Variables descriptions:
stress: 0=no high stress 1=high stress
gender: 0=male 1=female
curmar: 0=not married 1=married
cohab: 0=not cohabiting 2 = cohabiting


Commands:

logit stress gender curmar cohab

margins, at(gender==0 curmar==0 cohab==0) at(gender==0 curmar==0 cohab==1) ///
at(gender==0 curmar==1 cohab==0) at(gender==0 curmar==1 cohab==1) atmeans saving(f1, replace)
marginsplot

margins, at(gender==1 curmar==0 cohab==0) at(gender==1 curmar==0 cohab==1) ///
at(gender==1 curmar==1 cohab==0) at(gender==1 curmar==1 cohab==1) atmeans saving(f2, replace)
marginsplot

combomarginsplot f1 f2, noci

On the comboplot, it is separated by file. File 1 is on the lest and file 2 is on the right.

Thank you

[Nicholas Winter]

This is very old, but I just noticed it now. You don't actually ask a question, but I suspect your question is, "how can I get the plots overlaid in a different way?"

The answer to that is that you can get *almost* what you want with the -plotdimension()- option of marginsplot:

combomarginsplot f1 f2, noci plotdim(_filenumber)

The problem is that combomarginsplot has no way to know that the first at() in each call to margins correspond to each other, that the second two correspond to each other, and so on.

Note also that I think you can get hte same underlying margins results in one go, using Stata's factor notation, and then use -marginsplot- to get close to what I think you want:

logit stress i.gender i.curmar i.cohab
margins gender#curmar#cohab

marginsplot , by(cohab)