Publication-ready tables of margins with bootstrapped standard errors

[Devra Golbe]

I am having trouble creating a publication-ready table from the output of a margins command called from within a bootstrap program. What I want to do is present the margins and standard errors for each outcome side by side. I can do this without the bootstrap. For instance, using Ben Jann's estout (SJ14-2:st0085_2)

clear all
set more off
use http://www.stata-press.com/data/r15/nmihs.dta

mlogit bwgrp marital age highbp, baseoutcome(3)
estimates store one
foreach o in  1  2 {
         margins, dydx(*) predict(outcome(`o')) post
         estimates store m1_`o'
        estimates restore  one
    }
    
estout  m1_1 m1_2, cells (b (star fmt(3)) t(par fmt(2)))

results in

--------------------------------------------
                     m1_1            m1_2  
                      b/t             b/t  
--------------------------------------------
marital            -0.046***       -0.034***
                  (-5.91)         (-4.39)  
age                 0.001*          0.000  
                   (2.21)          (0.10)  
highbp              0.070***        0.066***
                   (5.32)          (4.93)  
--------------------------------------------

How can I get a similar table if I need to bootstrap the standard errors? The following code stores the results of the mlogit command in bs1, not the margins command; I believe that m2_1 and M2_2 contain the results of the last iteration.

program bs_con, eclass
mlogit bwgrp marital age highbp , baseoutcome(3)
estimates store bsts
foreach o in  1  2 {
         margins, dydx(*) predict(outcome(`o')) post
         estimates store m2_`o'
        estimates restore bsts
    }
end program

bootstrap, reps(5) : bs_con
estimates store bs1

On the other hand, this code saves the results I want, but not in the format I'd like for presentation.

program bs_con2, eclass
mlogit bwgrp marital age highbp , baseoutcome(3)
estimates store bsts
margins, dydx(*)  post
end program

bootstrap, reps(5) : bs_con2
estimates store bs2

. estout  bs2, cells (b (star fmt(3)) t(par fmt(2)))

----------------------------
                      bs2   
                      b/t   
----------------------------
marital                     
1._predict         -0.046***
                  (-6.13)   
2._predict         -0.034***
                  (-7.64)   
3._predict          0.081***
                   (7.39)   
----------------------------
age                         
1._predict          0.001*  
                   (2.45)   
2._predict          0.000   
                   (0.14)   
3._predict         -0.002*  
                  (-2.24)   
----------------------------
highbp                      
1._predict          0.070***
                   (7.94)   
2._predict          0.066***
                   (4.51)   
3._predict         -0.136***
                 (-11.96)   
----------------------------

Any advice would be much appreciated.

Thanks,
Devra

[Phil Bromiley]

Stata will let you copy and past results as tables. It may be less trouble to set up the table in your paper and copy and paste these results into the right columns than it would be to program it.

[Devra Golbe]

Thanks, Phil. I was hoping to avoid that, but perhaps it can't be helped.

Best,
Devra