Data.Reform: Reform the data

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

Description

Reform the data to a compatible form for ITT, HTE, SharpBound, and LongHudgens functions.

Usage

Data.Reform(Y, Z, A, C, X, seed=1)

Arguments

Y	N-by-1 vector containing the outcome of individuals where N is the number of individuals in the population.
Z	N-by-1 0-1 vector containing the treatment assignment where Z=1 indicates an individual was assigned to treatment.
A	N-by-1 0-1 vector containing the actual treatment usage of individuals where A=1 indicates an individual actually used treatment.
C	N-by-1 numeric vector containing cluster ID where individuals from the same cluster have the same C value.
X	N-by-p numeric matrix or data frame containing the pre-treatment covariates.
seed	(Optional) A random seed to generate the random noise used in SharpBound; see Section 4.2 of Park and Kang (2021) for details.

Value

Y	N-by-1 vector; same as input Y.
Z	N-by-1 vector; same as input Z.
A	N-by-1 vector; same as input A.
C	N-by-1 vector; same as input C.
X	N-by-p matrix or data frame; same as input X.
N	The number of individuals in the population.
J	The number of clusters in the population.
Class	J-by-1 vector of cluster ID.
nc	J-by-1 vector of cluster sizes.
n	N-by-1 vector of cluster sizes and the same as rep(nc,nc).
Zc	J-by-1 vector of treatment assignment and the simplified form of input Z.
m	The number of clusters that were assigned to treatment.
error	N-by-3 vector of the random noises and are used in SharpBound; see Section 4.2 of Park and Kang (2021) for details.

References

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)

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