We downloaded from the official website of ACIC and packaged these codes（Data Generator Processor, DGP） into a zip file: ACIC_2019_Generate_DGPs.zip
We generated the dataset using the method4 in the SG_Generate_High_Dim_Binary/generate_simEpilepsy.R which can be found in ACIC_2019_Generate_DGPs.zip:
#------------------------------------------------------------------
# mod 4: do treatment heterogeneity along with IVs.
#------------------------------------------------------------------
set.seed(40)
beta.A.mod4 <- runif(ncol(temp.mod2), -.1, .12) / apply(temp.mod2, 2, sd)
logitA.mod4 <- -.1 + temp.mod2 %*% beta.A.mod4
beta.Y.mod4 <- 2 * beta.A.mod4
beta.Y.mod4[1:5] <- 0 # create IVs
logit.drs.mod4 <- -1.8 + as.matrix(temp.mod2) %*% beta.Y.mod4+ as.matrix(d[,c(150, 160)]) %*% c(-.005, -0.02)
set.seed(21)
# true ATE in the population
psi0.mod4 <- mean(plogis( 2 + .01 * d[,160] + logit.drs.mod4) - plogis(logit.drs.mod4)) # 0.2916274
# Generate datasets
d.mod4 <- data.frame(Y= NA, A= NA, d)
set.seed(4)
niter <- 100
n.b <- 2000
for (i in mod4_files){
b <- sample(1:n, n.b, replace = TRUE)
d.mod4$A[b] <-rbinom(n.b, 1, plogis(logitA.mod4)[b])
d.mod4$Y[b]<- rbinom(n.b, 1, plogis(d.mod4$A * (2 + .01 * d[,160]) + logit.drs.mod4)[b])
write.csv(d.mod4[b,], file = paste0("epilepsyMod4", i , ".csv"), row.names = FALSE)
}