using margins to predict from 2 level categorical variable

[Martin Imelda Borg]

I have code a variable - categorical, which is called -severity- (excellent (1) - fair (2) - poor (3)) for a treatment variable (treatment = 1, treatment = 0 control)

I am trying to plot onto a graphs to predict the probability of the person having excellent or fair or poor outcomes for treatment variables.
However, I can't plot a line graph as the dependent variable is categorical, do you have any ideas how to plot the marginsplot graphs into 1 graph?

ologit severity treatment age gender [pw=weight]

margins, at(treatment=(0/1)) predict(outcome(1)) atmeans ///excellent
marginsplot

margins, at(treatment=(0/1)) predict(outcome(2)) atmeans ///fair
marginsplot

margins, at(treatment=(0/1)) predict(outcome(3)) atmeans ///poor
marginsplot

ologit everity treatment age gender [pw=weight]
predict probexcellent probfair probpoor 

twoway (line probexcellent treatment, lcolour(dkgreen) lpattern(solid)) (line probfair treatment, lcolour(dknavy) lpattern(solid))

[George Ford]

bar chart with 0/1 outcomes for each group?

[Martin Imelda Borg]

Thanks for this advice
I've generated this graph (but they're 4 separate bar charts) is there a way how to combine them all on the same x axis into 1 graph (excellent - poor - fair) and instead label where '1' as: Excellent - Fair - Poor with a common y axis ( probability)

If I'm not explaining myself well, I've edited the output in the desired picture.

ologit severity treatment age gender [pw=weight]

margins, at(treatment=(0/1)) predict(outcome(1)) atmeans ///excellent
marginsplot, plot(treatment) recast(bar) allxlabels by(treatment) name(m1, replace)

margins, at(treatment=(0/1)) predict(outcome(2)) atmeans ///fair
marginsplot, plot(treatment) recast(bar) allxlabels by(treatment) name(m2, replace)


margins, at(treatment=(0/1)) predict(outcome(3)) atmeans ///poor
marginsplot, plot(treatment) recast(bar) allxlabels by(treatment) name(m3, replace)

graph combine m1 m2 m3, ycommon name(g2, replace)

[Erik Ruzek]

Not ideal, but a start:

margins, at(treatment=(0/1)) predict(outcome(1)) predict(outcome(2)) predict(outcome(3)) atmeans 
marginsplot

[Martin Imelda Borg]

thanks for trying, but using that code - stata returns the error::

) required

[Erik Ruzek]

That's weird. I'm not sure why that is happening. The following, with the exact same code, works for me in Stata 16.1.

webuse nhanes2f, clear
ologit health i.female age c.age#c.age [pw=finalwgt]
margins , at(female=(0/1)) predict(outcome(1)) predict(outcome(2)) predict(outcome(3)) predict(outcome(4)) atmeans 
marginsplot,