tests/testthat/test_state_aak_als.R

In cvcqv: Coefficient of Variation (CV) with Confidence Intervals (CI)

context("state_aak_als")

# Infants weights (in grams) data, Abu-Shawiesh, Akyuz & Kibria (2019),
# Section 6.1 (n = 61).
infants_data <- c(
  4960, 5130, 4260, 5160, 4050, 5240, 4350, 4360, 3930, 4410,
  4610, 4102, 3530, 4550, 4460, 2940, 4160, 4110, 4410, 4800,
  5130, 3670, 4550, 4290, 5210, 4950, 5210, 3210, 4030, 3580,
  4360, 4360, 3920, 4050, 4630, 3756, 4382, 4586, 5336, 2828,
  4172, 4256, 4594, 4866, 4784, 4520, 5238, 4320, 5070, 5330,
  3836, 5916, 5010, 4344, 3496, 4148, 4044, 5192, 4368, 4180,
  5044
)

pmi_data <- c(
  5.5, 14.5, 6.0, 5.5, 5.3, 5.8, 11.0, 6.1, 7.0, 14.5,
  10.4, 4.6, 4.3, 7.2, 10.5, 6.5, 3.3, 7.0, 4.1, 6.2,
  10.4, 4.9
)

small_data <- c(
  0.2, 0.5, 1.1, 1.4, 1.8, 2.3, 2.5, 2.7, 3.5, 4.4,
  4.6, 5.4, 5.4, 5.7, 5.8, 5.9, 6.0, 6.6, 7.1, 7.9
)

test_that(
  desc = "AA&K-ALS matches paper Table 5 on infants data",
  {
    # Paper reports (0.1155, 0.1687).
    result <- CoefVarCI$new(
      x = infants_data, alpha = 0.05, digits = 6
    )$aak_als_ci()
    expect_equal(
      result$statistics$lower / 100, 0.1155,
      tolerance = 0.005
    )
    expect_equal(
      result$statistics$upper / 100, 0.1687,
      tolerance = 0.005
    )
  }
)

test_that(
  desc = "AA&K-ALS matches paper Table 6 on PMI data (wide tolerance)",
  {
    # Paper reports (0.3128, 0.6464). PMI is a small-n (22),
    # heavily-skewed sample where AA&K-ALS sits near the edge of
    # its three-term Taylor regime. We allow a 1 percentage-point
    # slack, mirroring the tolerance used by the Python tests in
    # pycvcqv/tests/test_aak_als.py.
    result <- CoefVarCI$new(
      x = pmi_data, alpha = 0.05, digits = 6
    )$aak_als_ci()
    expect_equal(
      result$statistics$lower / 100, 0.3128,
      tolerance = 0.01
    )
    expect_equal(
      result$statistics$upper / 100, 0.6464,
      tolerance = 0.01
    )
  }
)

test_that(
  desc = "AA&K-ALS bounds bracket the point estimate",
  {
    result <- CoefVarCI$new(
      x = small_data, digits = 6
    )$aak_als_ci()
    expect_lt(result$statistics$lower, result$statistics$est)
    expect_gt(result$statistics$upper, result$statistics$est)
  }
)

test_that(
  desc = "AA&K-ALS requires at least 4 observations",
  {
    expect_error(
      CoefVarCI$new(x = c(1.0, 2.0, 3.0))$aak_als_ci(),
      regexp = "(at least 4|G_2)"
    )
  }
)

test_that(
  desc = "cv_versatile aak_als matches the R6 method",
  {
    r6 <- CoefVarCI$new(
      x = small_data, alpha = 0.05, digits = 6
    )$aak_als_ci()
    proc <- cv_versatile(
      x = small_data, method = "aak_als", alpha = 0.05, digits = 6
    )
    expect_equal(
      r6$statistics$lower, proc$statistics$lower,
      tolerance = 0.0001
    )
    expect_equal(
      r6$statistics$upper, proc$statistics$upper,
      tolerance = 0.0001
    )
  }
)

test_that(
  desc = "AA&K methods are included in all_ci row names",
  {
    r6_all <- CoefVarCI$new(
      x = small_data, alpha = 0.05
    )$all_ci()
    rn <- rownames(r6_all$statistics)
    expect_true("aak_adj" %in% rn)
    expect_true("aak_ls" %in% rn)
    expect_true("aak_als" %in% rn)
  }
)