hte
provides functions for estimating heterogenous treatment effects.
Metalearners for estimating heterogeneous treatment effects using machine learning. PNAS 2019
The latest development version can be installed directly from Github using devtools:
if (!require("devtools")) install.packages("devtools")
devtools::install_github("soerenkuenzel/causalToolbox")
The package contains compiled code, and you must have a development environment to install the development version. (Use devtools::has_devel()
to check whether you do.) If no development environment exists, Windows users download and install Rtools and macOS users download and install Xcode.
For details please read our paper: https://arxiv.org/pdf/1706.03461.pdf This example will not execute because there is no data.
library(causalToolbox)
packageVersion("causalToolbox")
# create example data set
simulated_experiment <- simulate_causal_experiment(
ntrain = 1000,
ntest = 1000,
dim = 10,
setup = "complexTau",
testseed = 293901,
trainseed = 307017
)
feature_train <- simulated_experiment$feat_tr
w_train <- simulated_experiment$W_tr
yobs_train <- simulated_experiment$Yobs_tr
# create the hte object using honest Random Forests (RF)
xl_rf <- X_RF(feat = feature_train, tr = w_train, yobs = yobs_train)
# alternatively, use BART instead of honest Random Forests. If you are not going
# to be careful about hyperparemeter tuning, we suggest using BART.
xl_bart <- X_BART(feat = feature_train, tr = w_train, yobs = yobs_train)
# estimate the CATE
feature_test <- simulated_experiment$feat_te
cate_esti_rf <- EstimateCate(xl_rf, feature_test)
cate_esti_bart <- EstimateCate(xl_bart, feature_test)
# evaluate the performance
cate_true <- simulated_experiment$tau_te
mean((cate_esti_rf - cate_true) ^ 2)
mean((cate_esti_bart - cate_true) ^ 2)
# Create confidence intervals via bootstrapping. Note that honest Random
# Forests have theoretically valid CIs, BART does not. However, BART often
# performs well for prediction without extensive tuning. See our paper
# (https://arxiv.org/abs/1706.03461)
# do not run (takes a long time)
xl_ci_rf <- CateCI(xl_rf, feature_test, B = 500)
xl_ci_bart <- CateCI(xl_bart, feature_test, B = 500)