Command not found error of self design bootstrap

[Ning Wei]

Here is a snippet of my code, and the environment is Stata 16SE. I receive the error message as follows:

capture program drop myboot
program define myboot, rclass
    args i
    preserve
    bsample
    gen new7`i'id=id
    xtset new7`i'id qdate
    xtreg F`i'.y $bplinshock $bplinxlist, fe robust
    scalar bylinh`i'=\_b[$bplinshock]
    gen new8`i'id=id
    xtset new8`i'id qdate
    xtreg F`i'.g $bplinshock $bplinxlist, fe robust
    scalar bglinh`i' = \_b[$bplinshock]
    scalar sumliny = bylinh`i' + sumliny
    scalar sumling = bglinh`i' + sumling
    return scalar multlinh`i'=sumliny/sumling
    restore
end

forvalues i = 1/20 {

clear all
set obs 100
  
simulate multlinh`i'=r(multlinh`i'), reps(100): myboot `i'
sum
bstat, stat(multlinh`i'\_obs) n(100)
estat bootstrap, all
  }

  
 myboot command not found r(111);

[Joseph Coveney]

Maybe change

clear all

to

drop _all

I can't actually follow your code, but it looks as if even that change won't make it work.

I guess that this thread is a continuation of the one over in the Mata subforum? When you get your next error message, you can add to this thread rather than starting a third.

[Joro Kolev]

I also do not see what this program is supposed to do, but the particular error message is certainly caused by -clear all- as Joseph suggested. -clear all- among dropping many other things, drops also all defined programs. So by the time the program is called, it is not in the memory anymore.

help clear

[Ning Wei]

Really appreciate your help! Thank you so much~

[Ning Wei]

Hi! I am bothering you again. Very much thanks for the help. I modified my code following the recommended instructions. The main idea of the program is to carry out two xtreg regressions and derives the _b1[] and _b2[] in order to simulate the _b1[]/_b2[]. Unfortunately, I still receive the error message as follows:

save SmallOpenEconomy1.dta,replace
  
capture program drop myboot
program define myboot, rclass
    args i
    preserve
    use SmallOpenEconomy1.dta, clear
    gen new7`i'id=id
    gen new8`i'id=id
    bsample, cluster(id) idcluster(new7`i'id)
    xtset, clear
    xtset new7`i'id qdate
    xtreg F`i'.y $bplinshock $bplinxlist, fe robust
    scalar bylinh`i'= _b[$bplinshock]
    bsample, cluster(id) idcluster(new8`i'id)
    xtset, clear
    xtset new8`i'id qdate
    xtreg F`i'.g $bplinshock $bplinxlist, fe robust
    scalar bglinh`i' =  _b[$bplinshock]
    replace sumliny = bylinh`i' + sumliny
    replace sumling = bglinh`i' + sumling
    return scalar multlinh`i'=sumliny/sumling
    restore
end

forvalues i = 1/20 {

drop _all
set obs 100
  
simulate multlinh`i'=r(multlinh`i'), reps(100): myboot `i'
sum
bstat, stat(multlinh`i') n(100)
estat bootstrap, all
  }

repeated time values within panel
an error occurred when simulate executed myboot

[Joseph Coveney]

The main idea of the program is to carry out two xtreg regressions and derives the _b1[] and _b2[] in order to simulate the _b1[]/_b2[].

If that's what your objective is, then your code seems at least to me to go off the rails. (I'm still not sure what it is you're trying to accomplish.)

Regardless, from the looks of it, your current error is coming from your xtset command, because you're drawing a random sample with replacement just beforehand.

[Joro Kolev]

The problem might be coming from this

gen new7`i'id=id
bsample, cluster(id) idcluster(new7`i'id)

I do not know what this does.

I think the way how you are supposed to use this construct is, you have a variable id which identifies the cluster, idnew does not exist in your data, then you write

bsample, cluster(id) idcluster(idnew)

I tried and I see that if you already have in your data idnew, like you do, bsample does not generate an error. But whether this does what you think is a different story.

My guess is that the when a cluster gets resampled more than once, the way how you do it does not result in a unique idnew, and hence the repeated values in your bootstrap samples.

[Ning Wei]

Thank you so much!!!!!!! (*^▽^*)After the modification, I finally generate a satisfying output.

save SmallOpenEconomy1.dta,replace
  
capture program drop myboot
program define myboot, rclass
    args i
    preserve
    use SmallOpenEconomy1.dta, clear
    bsample, cluster(id) idcluster(idnew)
    xtset, clear
    xtset idnew qdate
    xtreg F`i'.y $bplinshock $bplinxlist, fe robust
    scalar bylinh`i'= _b[$bplinshock]
    xtreg F`i'.g $bplinshock $bplinxlist, fe robust
    scalar bglinh`i' =  _b[$bplinshock]
    replace sumliny = bylinh`i' + sumliny
    replace sumling = bglinh`i' + sumling
    return scalar multlinh`i'=sumliny/sumling
    restore
end

forvalues i = 1/20 {

drop _all
set obs 100
  
simulate multlinh`i'=r(multlinh`i'), reps(100): myboot `i'
sum
  }

This can perfectly work out the problem.

Simulations (100)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
..................................................    50
..................................................   100

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
   multlinh8 |        100    .5106296    .0709337   .3720216   .7186825
number of observations (_N) was 0, now 100