LTfunction: Determines the splitting process and split criterions

In AshwiniKV/TEHTree: TEHTree: Treatment effect heterogeneity tree

Description

This function generates the tree structure as a list function.

Usage

LTfunction(
  Y,
  Z,
  X,
  X.i,
  version = "prognostic",
  pval.thresh = 0.05,
  min.split.size = 10,
  sample = "single"
)

Arguments

Y	response vector
Z	treatment indicator
X	covariate matrix
X.i	covariate matrix - ids
version	type of score - default is the prognostic score
pval.thresh	Threshold of p-value
min.split.size	Minimum size to split
sample	Use a single or double sample - default is single. The double sample approach uses a separate sample to construct the tree and to estimate the effects. This can be modified as required.

Examples

set.seed(10)
N<-2000
numx <- 5
alpha <-0.8
theta<-0.8
beta<- c(1,.8,.6,.4,.2)
gamma <- 1
Z <- rep(c(0, 1), each = N/2)
sigma <- diag(numx)
X.i <- mvrnorm(N,mu=rep(0,numx),Sigma=sigma)
W <- Z * ifelse(X.i[,1] > 0, 1, 0)
mu <- alpha + theta*Z + X.i %*% beta + W * gamma
Y <- rnorm(N, mean=mu)

# Single sample approach
LT1<-LTfunction(Y, as.logical(Z), X.i, X.i)

# Double sample approach
Nused <- 200
subjects <- c(1:(Nused/2), 1001:(1000+Nused/2))
LT2<-LTfunction(Y[subjects], Z[subjects], X.i[subjects,], X.i[subjects,], sample = "double")

AshwiniKV/TEHTree documentation built on Sept. 15, 2021, 11:21 p.m.