[dindays]

Hi Richard,

Thank you for your suggestions.

I did the single outcome using probit and then for getting the predictions across grades I use mprobit. And I think I cannot compare the summation of the probabilities in each grade from the mprobit outcomes to the single outcome from probit. Am I correct?

So here is the code that I use:

probit student_smp ib4.year gender urban head age mom_educ conspc12 schoolage_children [fw=weind] if (age>=13&age<=15), r
margins year, atmeans

mprobit student_smp4 ib4.year gender urban head age mom_educ conspc12 schoolage_children [fw=weind] if (age>=13&age<=15), r
margins year, atmeans predict(outcome(1))
margins year, atmeans predict(outcome(2))
margins year, atmeans predict(outcome(3))

So I will get outcome from each grade accurately rather than using separated probit model for each.


However, now I want to make sure that what I am doing is correct. Do you think multinomial probit is correct rather than oprobit? I read on ordered probit where the dependent value has a meaningful sequential order. Do you think that would apply on my case here?

I attempted to use oprobit and using the same predicted probabilties of my multinomial probit here.

Grateful for your help.

Best,
Daim

[Richard Williams]

I am on the road with iPad only so it is hard to answer complicated questions! But I don't see how that corresponds with what I suggested. For each grade there should be a probit model limited to the cases at risk of entering that grade. I don't see how your code does that. I think the separate grade regressions limited to those at risk is superior to my earlier mprobit idea.

[dindays]

Dear Richard,

Thank you, sorry for bothering you a lot with my questions.

I will think about it again on which model that fits to my available data.

Thanks for a billion.

Best,
Daim

[dindays]

Hi Richard,

If I use this following command:
probit student_smp ib4.year gender urban head age mom_educ conspc12 schoolage_children [fw=weind] if (age>=13&age<=15), nocons

How to evaluate the predicted probabilities if I want year_1==1, year_2==0, year_3==0 and the rest of time dummies set to be zeros and the other explanatory variables set to be at their mean values? I mean what is the syntax to do that?

I tried several things using margins at(year_1=1, year_2=0, etc until year_11=0) atmeans

but stata keeps refusing this.

Grateful for your help.

Best,
Daim

[Richard Williams]

margins year, atmeans