How to randomly sampling and regression, and save the output 500 times, and draw kdensity of 500 coeffients?

[Eric Cheung]

for example, I have a sample of 2000 students from 50 schools.

I want to randomly drop 2 schools and run a regression. Repeat this process 500 times. (I can do this)

What I don't know is that how to save 500 outputs (or coefficient) of regression? And draw the kdensity of 500 coefficients?

Thank you very much!

[FernandoRios]

you can do a program that "wraps" the process to later simulate:

program my_sim, eclass
preserve
    *** Drop 2 schools
   regress y x 
restore
end

simulate, reps(500): my_sim
kdensity "coefficient of interest"

[Eric Cheung]

Thank you very much , I can understand your code! But I don't know how to save the coefficient of x?

[Eric Cheung]

program my_sim, eclass
use "D:\stata\xxxx.dta",clear
preserve
tempfile temp
bys school: keep if _n==1
sample 48,count
sort school
save `temp'
restore
merge m:1 school using `temp'
keep if _merge == 3
drop _merge
reg y x1 x2
end
simulate, reps(500): my_sim

but can not run the code " simulate, reps(500): my_sim"?
how to save the coefficient of x1?

[FernandoRios]

Because regress was the last command within the program, after you do simulate, it should store ALL the coefficients fo rALL variables.
Otherwise, you can do something like
matrix b = _b[x1]
ereturn post b

HTH

[Maarten Buis]

Some time ago I presented a tutorial on how to use simulate, which may be helpful: http://maartenbuis.nl/presentations/chicago08.html

[Eric Cheung]

Because regress was the last command within the program, after you do simulate, it should store ALL the coefficients fo rALL variables.
Otherwise, you can do something like
matrix b = _b[x1]
ereturn post b

HTH

Thank you very much! ! I succeeded.

[Eric Cheung]

Some time ago I presented a tutorial on how to use simulate, which may be helpful: http://maartenbuis.nl/presentations/chicago08.html

Thank you very much! I'm going to learn this.