build_MOTTE_forest_CO: buildForest

In boyiguo1/MOTTE.RF: What the Package Does (One Line, Title Case)

Description

The function to fitting tree/trees

Usage

build_MOTTE_forest_CO(
  x.b,
  x.diff.1,
  x.diff.2,
  y.diff.1,
  y.diff.2,
  nsplits = NULL,
  nodesize = 2 * (ncol(x.b) + 1),
  left.out = 0.1,
  ntree = ifelse(is.null(nsplits), 1, 200),
  nCore = ifelse(is.null(nsplits), 1, detectCores() - 1)
)

Arguments

x.b	Before treatment covariates, a n by p matrix
nsplits	The number of split condidate want to examine when constructing split rule
nodesize	Parameter to control the node size. When the number of observations in Node smaller than nodesize, stop splitting
left.out	left.out is ensure at least left.out*2 sample for either treated or untreated sample in the group
ntree	Number of trees want to construct. By default it is 1; however, when Random method used, recommand setting it as 200
nCore	The number of cores use for forest contruction when doing parallel computation
x.e	After treatment covariates, a n by p matrix
treat	Treatment received, a n by 1 vector
y.b	Before treatment outcomes, a n by q matrix
y.e	After treatment outcomes, a n by q matrix #@param method Method to use when choosing split value. Two options: "Exhaust" and "Random"

Value

a list of data.tree. Even when only one tree construct, it is a list containing the single tree

Examples

#' set.seed(1)
B <- create.B(10)
Z <- create.Z(10, 3)

tmp.dat <- sim_MOTTE_data( n.train = 500, n.test = 200,
p = 10, q = 3, ratio = 0.5,
B = B, Z = Z)

train.dat <- tmp.dat$train

x.b <- scale(train.dat$x.b, center = FALSE, scale = TRUE)
x.e <- scale(train.dat$x.e, center = FALSE, scale = TRUE)
y.b <- scale(train.dat$y.b, center = FALSE, scale = TRUE)
y.e <- scale(train.dat$y.e, center = FALSE, scale = TRUE)
treat <- train.dat$trt

# with(train.dat,
    build_MOTTE_forest(x.b, x.e, treat, y.b, y.e)
#)

boyiguo1/MOTTE.RF documentation built on June 14, 2020, 4:12 p.m.