tests/testthat/test-adaptive_iptw-missingAY.R

In drtmle: Doubly-Robust Nonparametric Estimation and Inference

library(drtmle)
library(SuperLearner)
library(np)
library(testthat)

# TO DO: Add tests for multiple treatment levels
context("Testing adaptive_iptw works")
test_that("adaptive_iptw works as expected with missing data", {
  set.seed(123456)
  n <- 200
  W <- data.frame(W1 = runif(n), W2 = rnorm(n))
  DeltaA <- 1 - rbinom(n, 1, plogis(-4 + W$W1 - W$W2))
  A <- rbinom(n, 1, plogis(W$W1 - W$W2))
  DeltaY <- 1 - rbinom(n, 1, plogis(-4 + W$W1 - A))
  Y <- rnorm(n, W$W1 * W$W2 * A, 2)
  Y[DeltaY == 0] <- NA
  A[DeltaA == 0] <- NA

  fit1 <- adaptive_iptw(
    W = W, A = A, Y = Y,
    a_0 = c(0, 1),
    glm_g = "W1 + W2",
    glm_Qr = "gn"
  )
  expect_true(all(!is.na(fit1$iptw_tmle$est)))
  expect_true(all(!is.na(fit1$iptw_tmle$cov)))
  expect_true(all(!is.na(fit1$iptw_os$est)))
  expect_true(all(!is.na(fit1$iptw_os$cov)))
  expect_true(all(!is.na(fit1$iptw$est)))
  expect_true(inherits(fit1, "adaptive_iptw"))
})