Hello everyone,

This is more of a clarification question as my code runs without any errors:

I am working with NHANES data with the following characteristics:


Code:
. svydescribe

Survey: Describing stage 1 sampling units

      pweight: wtint2yr
          VCE: linearized
  Single unit: centered
     Strata 1: sdmvstra
         SU 1: sdmvpsu
        FPC 1: <zero>
Let

outcome= outcome variable
cov1=covariate 1
cov2=covariate 2
cov3=covariate 3
cov4=covariate 4


I have the following simple two step estimation procedure:

Code:
capture noisily program drop twostep
global probitcovariates cov2 cov3 cov4
program twostep, rclass
args y x covariates
tempvar touse
gen byte `touse' = 1
*First Step
svy: probit `x' $probitcovariates
predict xd2h
gen phi2h = normalden(xd2h)
gen PHI2h = normal(xd2h)
gen gr2 = `x'*phi2h/PHI2h - (1 - `x')*phi2h/(1 - PHI2h)
gen `x'_gr2= `x'*gr2
*Second Step
svy:reg `y' `x' `covariates' `x'_gr2 gr2 
*Dropping Variables w/ gr
drop xd2h phi2h PHI2h gr2 `x'_gr2
end
Note that I am using svy: estimation command in both the steps.

I want to bootstrap my final estimates to account for the first step estimation. My code is

Code:
global covariates cov2 cov3
local bootreps = 100
bootstrap _b, reps(`bootreps') seed(123): twostep outcome cov1 "$covariates"
My code smoothly without any errors. However, since I am using survey data, I wanted to make sure that even though I am using svy at each step, do I still need
Code:
svy bootstrap
again in the final command?


Thank you so much for your time and consideration.