PAL.control: Control parameters for penalized A-learning

In ITRSelect: Variable Selection for Optimal Individualized Dynamic Treatment Regime

Description

Parameters that control fitting of penalized A-learning.

Usage

PAL.control(pi1.est = NULL, pi2.est = NULL, h1.est = NULL, h2.est = NULL, kappa = NULL, 
  penalty = 'SCAD')

Arguments

pi1.est	Estimated propentisy score at the first stage. By default, a penalized logistic regression model is fitted to estimate the propensity score.
pi2.est	Estimated propentisy score at the second stage. By default, a penalized logistic regression model is fitted to estimate the propensity score.
h1.est	Estimated baseline function at the first stage. By default, a penalized linear regression model is fitted to estimate the baseline function.
h2.est	Estimated baseline function at the second stage. By default, a penalized linear regression model is fitted to estimate the baseline function.
kappa	The model complexity penalty used in the information criteria. By default, kappa=1 if BIC or CIC is used and kappa=4 if VIC is used.
penalty	The penalty to be applied to the propensity score and baseline model. Either "MCP", "SCAD" (the default), or "lasso".

Value

A list with the arguments specified.

See Also

PAL, PAL.fit

Examples

set.seed(12345)
n <- 200
p <- 200
X <- matrix(rnorm(n*p), nrow=n, ncol=p)
A <- rbinom(n, 1, 0.5)
CX <- (X[,1] + X[,2])
h <- 1 + X[,1] * X[,3]
Y <- h + A*CX + 0.5*rnorm(n)
result <- SAS(Y~X|A, pi1.est=0.5)

ITRSelect documentation built on May 1, 2019, 10:56 p.m.