Adaptive Targeted Minimum Loss-Based Estimation. This package uses adaptive targeted minimum loss-based estimation to estimate the average treatment effect from combined randomized trial and real-world data.
Authors: Sky Qiu, Lars van der Laan Mark van der Laan,
If you encounter any bugs or have any specific feature requests, please file an issue.
library(atmle)
library(EScvtmle)
options(sl3.verbose = TRUE)
`%+%` <- function(a, b) paste0(a, b)
set.seed(82379)
# simulate data
n <- 2000
S <- rbinom(n, 1, 0.2)
W1 <- rnorm(n); W2 <- rnorm(n); W3 <- rnorm(n)
A <- numeric(n)
g_rct <- 0.67
A[S == 1] <- rbinom(sum(S), 1, g_rct)
A[S == 0] <- rbinom(n-sum(S), 1, plogis(0.5*W1[S == 0]))
UY <- rnorm(n, 0, 1)
Y <- 2.5+0.9*W1+1.1*W2+2.7*W3+1.5*A+UY+(1-S)*(0.2+0.1*W1*(1-A))
data <- data.frame(S, W1, W2, A, Y)
true_ate <- 1.5
# run A-TMLE
res <- atmle(data,
S_node = c(1),
W_node = c(2, 3),
A_node = 4,
Y_node = 5,
controls_only = FALSE,
family = "gaussian",
atmle_pooled = TRUE,
g_rct = 0.67,
verbose = FALSE)
print("A-TMLE ATE estimate: " %+% round(res$est, 2) %+%
" (" %+% round(res$lower, 2) %+% ", " %+% round(res$upper, 2) %+% ")")
#> [1] "A-TMLE ATE estimate: 1.56 (1.3, 1.82)"
print("True ATE: " %+% true_ate)
#> [1] "True ATE: 1.5"
# compared to ES-CVTMLE
escvtmle_res <- ES.cvtmle(txinrwd = TRUE,
data = data,
study = "S",
covariates = c("W1", "W2"),
treatment_var = "A",
treatment = 1,
outcome = "Y",
pRCT = g_rct,
family = "gaussian",
Q.SL.library = c("SL.glm"),
g.SL.library = c("SL.glm"),
Q.discreteSL = TRUE,
g.discreteSL = TRUE,
V = 5)
print("ES-CVTMLE ATE estimate: " %+% round(escvtmle_res$ATE$b2v, 2) %+%
" (" %+% round(escvtmle_res$CI$b2v[1], 2) %+%
", " %+% round(escvtmle_res$CI$b2v[2], 2) %+% ")")
#> [1] "ES-CVTMLE ATE estimate: 1.3 (0.82, 1.72)"
print("True ATE: " %+% true_ate)
#> [1] "True ATE: 1.5"© 2023 Sky Qiu, Lars van der Laan Mark van der Laan,
The contents of this repository are distributed under the GPL-3 license.
See file LICENSE for details.