Marginal Fixed Effects on Ordered Probit

[Anna O'Donnell]

I am running an ordered probit model and am looking to calculate the marginal effects using the mfx post estimation command.

I am trying to get the estimated probability of the outcome 8 or greater. I am running the following command mfx compute, predict(outcome(8)) but I think the output provided is the estimated probability for outcome 8 ONLY. How do I get outcome for 8 and higher. My dependent variable ranges from 0-10.

I tried the following mfx compute, predict(outcome(>=8)) and this did not work.


Many thanks,
Anna

[Richard Williams]

First off, unless you are using an ancient version of stata, don't use mfx. Use margins instead.

Second, check out Nick Winter's -combomarginsplot- command, available from SSC. He shows some neat tricks I was not aware of for combining categories in an oprobit or ologit. Even if you don't use his program, you may be able to adapt his examples. In particular, I am guessing that you can modify this example that he gives:

        . sysuse auto
        . oprobit rep78 i.foreign mpg price weight
        . margins foreign, at(mpg=(10(5)50)) expression(predict(outcome(1))+predict(outcome(2))) saving(file4, replace)
        . margins foreign, at(mpg=(10(5)50)) expression(predict(outcome(4))+predict(outcome(5))) saving(file5, replace)

        . combomarginsplot file4 file5 , labels("Outcomes 1&2" "Outcomes 4&5") noci file1opts(pstyle(p1)) file2opts(pstyle(p2)) lplot1(mfcolor(white))