Hi,

I am wondering if someone can help me, as have been having some issues with plotting a figure to match the results from a post estimation margins command.

I have the following model:

regress y i.sex##c.age, vce(cluster studyid)

I am accounting for the fact that results are from 17 different studies (study id).

From this model, I obtain a significant interaction effect.

I then want to look at the outcome (y) at all levels of age, as therefore run:

margins sex, at(xcage=(7(1)18)), vce(unconditional)

(which also adjust the SE for the 17 clusters)

I then obtain a table with margins and SE that I would like to plot with:

marginsplot, horizontal recast(scatter) xline(119, lcolor(red))

However, the CIs for each age (by sex) overlap significantly (i.e. I was expecting sex differences at each age as I ran):

margins, dydx(sex) at(xcage=(7(1)18)) vce(unconditional)

...which shows most male/female differences are significant at the <0.001 level.

Therefore, my question is, why do the adjusted means (or the unconditional dy/dx / standard errors) do not match up to the marginal effects command?

Thank you in advance,
Lucy