tests/testthat/test_tauStarTest.R

In TauStar: Efficient Computation and Testing of the Bergsma-Dassios Sign Covariance

library(TauStar)
context("Testing the tauStarTest functionality.")

test_that("tauStarTest with continuous data.", {
  set.seed(238)
  sims <- 5
  n <- 100
  for (i in 1:sims) {
    x <- rnorm(n)
    y <- rnorm(n)
    tval <- tStar(x, y)
    a <- tauStarTest(x, y, mode = "continuous")
    expect_equal(a$x, x)
    expect_equal(a$y, y)
    expect_equal(a$tStar, tval)
    expect_equal(a$pVal, 1 - pHoeffInd(n * tval))
    expect_equal(a$mode, "continuous")
  }

  x <- rnorm(n)
  y <- rnorm(n)
  tval <- tStar(x, y)
  a <- tauStarTest(x, y)
  expect_equal(a$x, x)
  expect_equal(a$y, y)
  expect_equal(a$tStar, tval)
  expect_equal(a$pVal, 1 - pHoeffInd(n * tval))
  expect_equal(a$mode, "continuous")

  expect_error(tauStarTest(rnorm(10), rnorm(9)))
})

test_that("tauStarTest with discrete data.", {
  set.seed(238)
  sims <- 10
  n <- 2000
  for (i in 1:sims) {
    p <- runif(sample(1:6, 1))
    p <- p / sum(p)
    q <- runif(sample(1:6, 1))
    q <- q / sum(q)
    x <- sample(length(p), n, replace = TRUE, prob = p)
    y <- sample(length(q), n, replace = TRUE, prob = q)
    tval <- tStar(x, y)
    a <- tauStarTest(x, y, mode = "discrete")
    expect_equal(a$x, x)
    expect_equal(a$y, y)
    expect_equal(a$tStar, tval)
    expect_true(abs(a$pVal - (1 - pDisHoeffInd(n * tval, p, q))) <= 10^-2)
    expect_equal(a$mode, "discrete")
  }

  p <- runif(sample(1:6, 1))
  p <- p / sum(p)
  q <- runif(sample(1:6, 1))
  q <- q / sum(q)
  x <- sample(length(p), n, replace = TRUE, prob = p)
  y <- sample(length(q), n, replace = TRUE, prob = q)
  tval <- tStar(x, y)
  a <- tauStarTest(x, y)
  expect_equal(a$x, x)
  expect_equal(a$y, y)
  expect_equal(a$tStar, tval)
  expect_true(abs(a$pVal - (1 - pDisHoeffInd(n * tval, p, q))) <= 10^-2)
  expect_equal(a$mode, "discrete")

  expect_error(tauStarTest(rnorm(10), rnorm(9), mode = "discrete"))
})

test_that("tauStarTest with mixed data.", {
  suppressWarnings(RNGversion("3.5.0"))
  set.seed(238)
  sims <- 10
  n <- 200
  for (i in 1:sims) {
    p <- runif(sample(1:6, 1))
    p <- p / sum(p)
    if (sample(c(TRUE, FALSE), 1)) {
      x <- sample(length(p), n, replace = TRUE, prob = p)
      y <- rnorm(n)
    } else {
      y <- sample(length(p), n, replace = TRUE, prob = p)
      x <- rnorm(n)
    }
    tval <- tStar(x, y)
    a <- tauStarTest(x, y, mode = "mixed")
    expect_equal(a$x, x)
    expect_equal(a$y, y)
    expect_equal(a$tStar, tval)
    expect_true(abs(a$pVal - (1 - pMixHoeffInd(n * tval, p))) <= 10^-2)
    expect_equal(a$mode, "mixed")
  }

  p <- runif(sample(1:6, 1))
  p <- p / sum(p)
  x <- sample(length(p), n, replace = TRUE, prob = p)
  y <- rnorm(n)
  tval <- tStar(x, y)
  a <- tauStarTest(x, y, mode = "auto")
  expect_equal(a$x, x)
  expect_equal(a$y, y)
  expect_equal(a$tStar, tval)
  expect_true(abs(a$pVal - (1 - pMixHoeffInd(n * tval, p))) <= 10^-2)
  expect_equal(a$mode, "mixed")

  x <- rnorm(n)
  expect_warning(tauStarTest(x, y, mode = "mixed"))

  expect_error(tauStarTest(rnorm(10), rnorm(9), mode = "mixed"))
  expect_error(tauStarTest(c(1, 2, 3, 4, 1), c(1, 2, 3, 4, 1), mode = "mixed"))
})

test_that("tauStarTest with a permutation test.", {
  set.seed(238)
  sims <- 10
  n <- 50
  resamples <- 123
  for (i in 1:sims) {
    x <- rnorm(n)
    y <- rnorm(n)
    tval <- tStar(x, y)
    a <- tauStarTest(x, y, mode = "permutation", resamples = resamples)
    expect_equal(a$x, x)
    expect_equal(a$y, y)
    expect_equal(a$tStar, tval)
    expect_true(!is.null(a$pVal))
    expect_equal(a$resamples, resamples)
    expect_equal(a$mode, "permutation")
  }

  expect_error(tauStarTest(rnorm(10), rnorm(9), mode = "permutation"))
})