[George Ford]

Since its regression based, could suest be used, assuming you can extract the final regressions from the ado file? Might have to add an eststo in the ado (maybe after it runs, but it may be rewritten by subsequent calculations). Then there's the weights (if used) which could be problematic in suest.

[Julie Martin]

Thanks a lot! I'll start with the one covariate case.

[Amani Shayo]

Hello everyone, I am trying to apply JWIDID in a repeated cross-section dataset. But when i run the command it gives me this error:


. jwdid attendnursy , ivar(statefip) time(year) gvar(gvar) never
WARNING: Singleton observations not dropped; statistical significance is biased (link)
st_data_pool(): 3900 unable to allocate real <tmp>[46261503,20]
FixedEffects::partial_out(): - function returned error
<istmt>: - function returned error
r(3900);

.
Does it mean i need to use another computer with enough memory or there is a way to deal with this case.

Thanks in advance

[FernandoRios]

Looks like that
perhaps your tvar and gvar are not correctly defined

[Amani Shayo]

Looks like that
perhaps your tvar and gvar are not correctly defined

Actually, I have tried with another computer and i ended up with the same error,
this is how i have generated my variables
***generating a treat variable***
gen treat=1 if statefip==6 & year>=2015
replace treat=1 if statefip==8 & year>=2015
replace treat=1 if statefip==9 & year>=2015
replace treat=1 if statefip==10 & year>=2016
replace treat=1 if statefip==15 & year>=2016
replace treat=1 if statefip==17 & year>=2014
replace treat=1 if statefip==24 & year>=2014
replace treat=1 if statefip==32 & year>=2014
replace treat=1 if statefip==34 & year>=2020
replace treat=1 if statefip==35 & year>=2003
replace treat=1 if statefip==36 & year>=2019
replace treat=1 if statefip==41 & year>=2019
replace treat=1 if statefip==49 & year>=2005
replace treat=1 if statefip==50 & year>=2014
replace treat=1 if statefip==51 & year>=2021
replace treat=1 if statefip==53 & year>=1993
replace treat=1 if statefip==11 & year>=2014
replace treat=0 if treat==.

//generate first trat
gen first_treat=0
replace first_treat =2015 if statefip==6
replace first_treat=2015 if statefip==8
replace first_treat=2015 if statefip==9
replace first_treat=2016 if statefip==10
replace first_treat=2016 if statefip==15
replace first_treat=2014 if statefip==17
replace first_treat=2014 if statefip==24
replace first_treat=2014 if statefip==32
replace first_treat=2014 if statefip==50
replace first_treat=2014 if statefip==11
replace first_treat=2014 if statefip==35
replace first_treat=2014 if statefip==49

***generate individuals who are likely undocumented eqn 2***
gen likely_undoc=1 if citizen==3 & hispan>=1 & hispan<=4 & age>=22 & age<=45 & educ>=0 & educ<=6
replace likely_undoc=0 if likely_undoc==.

***Identifyig number of undocumented immigrants***
egen number_undoc=total(likely_undoc==1), by(serial)

***Generate undocumented household***
gen undoc_hh=1 if number_undoc>=1
replace undoc_hh=0 if number_undoc==0


***generate attend kindergarten***
gen attendkinder=1 if gradeatt==2
replace attendkinder=0 if attendkinder==.

***generate attend nursery***
gen attendnursy=1 if gradeatt==1
replace attendnursy=0 if attendnursy==.

***primaryattend***
gen primaryattend=1 if gradeatt>=3 & gradeatt<=4
replace primaryattend=0 if primaryattend==.

//new way
gen trt = (first_treat<=year)*(first_treat>0)
***gen state = int(countyreal/1000)
set seed 1
sample 90

***to generate gvar***
egen gvar = csgvar(trt), ivar(statefip) tvar(year)

tab year gvar
//JWDID I
jwdid attendnursy , ivar(statefip) time(year) gvar(gvar) never
jwdid attendkinder , ivar(statefip) time(year) gvar(gvar) never
jwdid primaryattend , ivar(statefip) time(year) gvar(gvar) never

//JWDID II
jwdid attendnursy , cluster(statefip) time(year) gvar(gvar) never group
jwdid attendkinder , cluster(statefip) time(year) gvar(gvar) never group
jwdid primaryattend , cluster(statefip) time(year) gvar(gvar) never group

My concern is on the way to set JWDID to condition my dataset on the undocumented immigrants. please if there will be a better way will be highly appriciated