tests/testthat/test-ci.R

In qte: Quantile Treatment Effects

# =============================================================================
# Title: Fast smoke tests — unconfoundedness QTE estimator
# Description: Regression checks for unc_qte (Firpo 2007 IPW estimator).
#   Uses a low biters count to keep the test fast. Reference ATE from first
#   verified run. The large negative value reflects the PSID observational
#   comparison group — a regression check that estimates stay stable, not a
#   causal claim.
# Author: Brant Callaway
# Last update: 2026-05-22
# Date created: 2026-05-18
# =============================================================================

# Smoke test: no covariates (simple quantile differences with bootstrap SE)
test_that("unc_qte returns correct structure and stable ATE", {
  set.seed(42)
  cq1 <- unc_qte(
    yname = "re78", dname = "treat",
    data  = lalonde.psid,
    probs = seq(0.25, 0.75, 0.25),
    biters = 20
  )
  expect_s3_class(cq1, "QTE")
  expect_length(cq1$qte, 3)
  expect_true(is.numeric(cq1$ate))
  expect_false(anyNA(cq1$qte))
  expect_equal(cq1$ate, -15204.78, tolerance = 1e-2)
})

# Smoke test: with covariates (propensity score reweighting); checks pscore fitted
test_that("unc_qte with xformla returns valid IPW estimates", {
  set.seed(42)
  cq2 <- unc_qte(
    yname   = "re78", dname = "treat",
    data    = lalonde.psid,
    xformla = ~ age + education + black + hispanic + married + nodegree,
    probs   = seq(0.25, 0.75, 0.25),
    biters  = 20
  )
  expect_s3_class(cq2, "QTE")
  expect_length(cq2$qte, 3)
  expect_false(anyNA(cq2$qte))
  expect_true(!is.null(cq2$pscore.reg))
  expect_equal(cq2$ate, -13115.59, tolerance = 1e-2)
})