Saving margins estimates with bootstrap se after Two Part models

[Alon Pertzikovitz]

Hi all,
I am estimating a two-part model (twopm) and want to store the margins estimates with the correct bootstrap standard errors - to later plot several models combined in a combomarginsplot.
Here is an example of what I am trying to do, what should I change in my code so it will store it correctly?:

program define dydx_boot, eclass
qui twopm y i.age if female==0, firstpart(logit) second part(regress, log) vce(cluster id)
qui margins, at(age=(21(1)42)) predict(duan) post nose
end

bootstrap _b, saving(male, replace) seed(14) reps(1000):dydx_boot


capture program drop dydx_boot

program define dydx_boot, eclass
qui twopm y i.age if female==1, firstpart(logic) second part(regress, log) vce(cluster id)
qui margins, at(age=(21(1)42)) predict(duan) post nose
end

bootstrap _b, saving(female, replace) seed(14) reps(1000):dydx_boot


combomarginsplot male female

Much appreciated,
Alon

[Andrew Musau]

Once the bootstrap command has executed, the estimates should be available in memory. So just store them.

bootstrap _b, saving(male, replace) seed(14) reps(1000):dydx_boot
estimates store male

and

bootstrap _b, saving(female, replace) seed(14) reps(1000):dydx_boot
estimates store female

[Alon Pertzikovitz]

Thank you Andrew,

Maybe my usage of the word "store" in the question was misleading.
I want to save the margins results after bootstrap as a .dta
The code I posted, saves the 1000 bootstrap replications.

[Andrew Musau]

The estimates are from the bootstrap command, so there is no option to do what margins does. You can use marginsplot following bootstrap (or using the stored estimates) or construct the dataset with these estimates.