Bound.Intersect: Sharpening Bounds of the Compliance Group ITT Effects Using...

In qkrcks0218/CRTBound: Analysis of Cluster Randomized Trials for Intent-to-Treat Effects and Network Effects

Description

Sharpening bounds of the compliance group ITT effects using the approaches in Grilli and Mealli (2008), Long and Hudgens (2013), and Park and Kang (2021).

Usage

Bound.Intersect(SharpBound.Object, LongHudgens.Object, level=0.95)

Arguments

SharpBound.Object	An object obtained from SharpBound function.
LongHudgens.Object	An object obtained from LongHudgens function.
level	(Optional) The confidence level of the confidence interval. The default option is 0.95.

Value

Bound	Bound estimates of the compliance group ITT effects. When SharpBound.Object and LongHudgegns.Object have the true bounds, the true bounds are availab.e
BootCIBound	Confidence intervals of the bounds of the compliance group ITT effects.

References

Grilli Leonardo & Mealli Fabrizia (2008) Nonparametric Bounds on the Causal Effect of University Studies on Job Opportunities Using Principal Stratification. Journal of Educational and Behavioral Statistics [Link]

Dustin M. Long & Michael G. Hudgens (2013) Sharpening Bounds on Principal Effects with Covariates. Biometrics [Link]

Chan Park & Hyunseung Kang (2021+) Assumption-Lean Analysis of Cluster Randomized Trials in Infectious Diseases for Intent-to-Treat Effects and Network Effects, Journal of the American Statistical Association [Link]

Examples

#################################
# Generate Population
#################################

J <- 100 ; m <- 60
nc <- sample(c(3,4,5),J,replace=T)
N <- sum(nc)
C <- rep(1:J,nc)
n <- rep(nc,nc)
X1 <- rnorm(N)
X2 <- rbinom(N,1,0.3)
X3 <- rep(rnorm(J),nc)
A0 <- rbinom(N,1,logistic(-2+2*X3))
A1 <- apply(cbind(A0,rbinom(N,1,logistic(-2+3*X1+3*X2+2*X3))),1,max)
OtherA1 <- (rep(aggregate(A1~C,FUN="sum")[,2],nc)-A1)/(n-1)
NT <- which(A1==0 & A0==0)
AT <- which(A1==1 & A0==1)
CO <- which(A1==1 & A0==0)
Y0 <- Y1 <- rep(0,N)
for(jj in NT){
    Y0[jj] <- rbinom(1,1,logistic(-2+2*OtherA1[jj]))
    Y1[jj] <- max(Y0[jj],rbinom(1,1,logistic(-2+2*OtherA1[jj])))
}
for(jj in AT){
    Y0[jj] <- rbinom(1,1,logistic(2+X1[jj]+X2[jj]))
    Y1[jj] <- max(Y0[jj],rbinom(1,1,logistic(2+X1[jj]+X2[jj])))
}
for(jj in CO){
    Y0[jj] <- rbinom(1,1,logistic(-2+2*OtherA1[jj]))
    Y1[jj] <- max(Y0[jj],rbinom(1,1,logistic(2+X1[jj]+X2[jj])))
}
X <- cbind(1,X1,X2,X3,n)

Zc <- rep(0,J)
Zc[sort(sample(J,m))] <- 1
Z <- rep(Zc,nc)
A <- Z*A1 + (1-Z)*A0
Y <- Z*Y1 + (1-Z)*Y0

#################################
# Reform the Data
#################################

Reformed.Data <- Data.Reform(Y,Z,A,C,X,seed=1)
Simulated.Data <- Simulation.Reform(Y0,Y1,Z,A0,A1,C,X,seed=1)

#################################
# Bounds
#################################

Bound1 <- SharpBound(Reformed.Data,paraC=1:5,method="Logistic",CIcalc=TRUE,SSsize=100,level=0.95,seed=1,Input.Type="Data")
Bound2 <- LongHudgens(Reformed.Data,paraC=c(3),CIcalc=TRUE,SSsize=100,level=0.95,seed=1,Input.Type="Data")
Bound3 <- Bound.Intersect(Bound1,Bound2,level=0.95,Input.Type="Data")

Bound1.Sim <- SharpBound(Simulated.Data,paraC=1:5,method="Logistic",CIcalc=TRUE,SSsize=100,level=0.95,seed=1,Input.Type="Sim")
Bound2.Sim <- LongHudgens(Simulated.Data,paraC=c(3),CIcalc=TRUE,SSsize=100,level=0.95,seed=1,Input.Type="Sim")
Bound3.Sim <- Bound.Intersect(Bound1.Sim,Bound2.Sim,level=0.95,Input.Type="Sim")

qkrcks0218/CRTBound documentation built on Dec. 22, 2021, 10:56 a.m.