ipw.ate: Estimate ATE via inverse propensity weighting

In swager/balanceHD: Approximately balanced estimation of average treatment effects in high dimensions.

Description

Estimate ATE via inverse propensity weighting

Usage

ipw.ate(X, Y, W, target.pop = c(0, 1), eps.threshold = 1/20,
  fit.method = c("elnet", "none"), alpha.fit = 0.9,
  prop.method = c("elnet", "randomforest"), alpha.prop = 0.5,
  prop.weighted.fit = FALSE, targeting.method = c("AIPW", "TMLE"),
  estimate.se = FALSE)

Arguments

X	the input features
Y	the observed responses
W	treatment/control assignment, coded as 0/1
target.pop	which population should the treatment effect be estimated for? (0, 1): average treatment effect for everyone 0: average treatment effect for controls 1: average treatment effect for treated
eps.threshold	cap on the estimated propensities
fit.method	the method used to fit mu(x, w) = E[Y | X = x, W = w]
alpha.fit	tuning paramter for glmnet in the mu model
prop.method	the method used to fit e(x) = P[W = 1 | X]
alpha.prop	tuning paramter for glmnet in the propsenity model
prop.weighted.fit	whether propensity weights should be used to as sample weights in outcome fit
targeting.method	how to combine the outcome and propensity model fits.

Value

ATE estimate

swager/balanceHD documentation built on Aug. 10, 2021, 1:54 a.m.