Dear Stata Users,

I am computing a multinomial logit and its predicted probabilities - command margins- as follows:

PHP Code:
svy: mlogit uniontype2  i.durationtype  i.employ2 ,  level(90)  rrr base 
PHP Code:
 margins, level(90) pr(out(1)) pr(out(2)) pr(out(0)) vsquish vce(unconditional) force atmeans post at(employ2=(1(1)4)) 
I am using marginsplot (mplotoffset for being precise) to graph the predicted probabilities with specific values of employ2.

PHP Code:
mplotoffset ,   xdim(_outcome)  legend( label(1 "E/E") label(2 "M/U,F/E") label(3 "M/E,F/U") label(4 "M/U,F/U") size(small) ring(5) symxsize(*0.5) row(2)) ytitle("Probability", height(10)) title("Employment (a)") offset(.1) recast(scatter)  plotregion(lcol(black))  xlab(,labsize(small) )  xtitle("") plot2opts(msymbol(D)   fcol(black) fintensity(30)) plot1opts(  fcol(black) fintensity(15) ) ci2opts(recast(rcap)) ci1opts(recast(rcap) ) plot3opts(msymbol(T)   fcol(black) fintensity(50)) plot4opts(msymbol(S)   fcol(black) fintensity(70)) xlab(1 "Continuing coh" 2 "Dissolution" 3"Marriage", labsize(medsmall)) ci4opts(recast(rcap) ) ci3opts(recast(rcap)  ) ylab(0(.1)1) 


Outcome 0 ("continuing coh") has a much higher predicted probability than the other two. Is there the possibility of introducing a second axis for outcome 0 and then allow a smoother comparison of trends for each outcome?



Array


Thank you very much.

- Lydia