ROWL: ROWL, a function for OWL and RWL.

In sambiostat/RICWL:

Description

OWL and RWL are benchmark models in our simulation study

Usage

ROWL(
  H,
  A,
  R2,
  pi = rep(1, n),
  pentype = "lasso",
  kernel = "linear",
  residual = TRUE,
  sigma = c(0.03, 0.05, 0.07),
  clinear = 2^(-2:2),
  m = 4,
  e = 1e-05
)

Arguments

H:	n by P feature matrix
A:	treatment, takes value -1 and +1 with size n
R2:	outcome or residual vector with length n
pi:	propensity score with length n
pentype:	penalty in the residual estimation process, useful or RWL only
kernel:	kernel used in SVM
residual:	True or False, if True then RWL is deployed, if FALSE then OWL is deployed. Default is True.
sigma:	hyper-parameter in SVM rbf kernel
clinear:	hyper-parameter in SVM rbf kernel
m:	number of fold for cross validation in parameter tunning
e:	least tolerated error

Examples

n = 200
p = 5
H = matrix(rnorm(n*p), nrow=n)
A = rbinom(n=n, size=1, prob=0.5)*2 - 1
R2 = 4*H[,1]^2 + A*(H[,2]^2 + H[,4]^2 < 1.5)
fit = ROWL(H, A, R2)
optimalITR_pred = predict(fit, H)

sambiostat/RICWL documentation built on Dec. 31, 2020, 3:17 a.m.