Replicating RDROBUST

[Lindsey Buck]

Hi all,

I am trying to replicate the results you get from rdrobust with the below code. You can at first see how I ran rdrobust with my variable "distp_offneg" where 0 is the cutoff. I then tried creating a couple of variables: an instrment indicating when my variable is >0 (above the cutoff), an instrument times my variable disp_offneg, and then a triangular weight. I tried to run a regression with these in it, and the same covariates, hoping to replicate the results but it didn't work out. Can anyone help me figure out why?

foreach v of varlist welfare {
rdrobust `v' distp_offneg if agech>=25 & agech<=30 & male==1, p(1) bwselect(mserd) covs(lincomeb02_1 fsize02_1  fsize_incomeb02 bweight_1 lincomeb4_1 fsize4_1 fsize_incomeb4 father_figure4 black state4) vce(cluster cluster4)
outreg2 using table5_2530.xls
local leftbw= e(N_h_l) 
local rightbw= e(N_h_r) 
gen normscore`v' = distp_offneg
gen instrument`v' = (normscore`v'>0)
gen normscore_r`v' = normscore`v'*instrument`v'
gen transformed`v' = normscore`v'/(`leftbw' )
replace transformed`v' = normscore`v'/(`rightbw') if normscore`v'>0
gen inbandwidth`v' = normscore`v'>-`leftbw' & normscore`v'< `rightbw'
gen triwt`v' = (1-abs(transformed`v'))*inbandwidth`v'
reg `v' distp_offneg normscore_r`v' instrument`v' lincomeb02_1 fsize02_1  fsize_incomeb02 bweight_1 lincomeb4_1 fsize4_1 fsize_incomeb4 father_figure4 black state4 if age>=25 & age<=30 & male==1 [pw=triwt`v'], vce(cluster cluster4)
outreg2 using table5_2530.xls, addstat("control mean", mean)
}

[Fei Wang]

Here is another replication of sharp RD with triangular kernel: https://www.statalist.org/forums/for...ear-regression. Your code looks fine except that leftbw and rightbw should be e(h_l) and e(h_r), respectively.