tests_to_process/test-quantile_algorithm.R

In vincenzocoia/distplyr: Manipulation of Univariate Distributions

d <- mix(dst_unif(0, 5), dst_empirical(c(2, 3, 3, 4, 6)))
# plot(d, "cdf", from = -1, to = 7, n = 1001)
cdf <- get_cdf(d)
tau <- c(0.1, 0.15, 0.2, 0.25, 0.3, 0.35, 0.4, 0.5, NA, NaN, 0.55, 0.6, 0.9, 1)
ans <- c(1, 1.5, 2, 2, 2, 2.5, 3, 3, NA, NaN, 3, 3, 5, 6)
tau_left <- c(-Inf, -100, 0)
tau_right <- c(100, Inf)
ans_left <- c(-Inf, -Inf, -Inf)
ans_right <- c(Inf, Inf)
set.seed(1)
permute <- sample(1:length(tau))

test_that("quantiles by inversion of cdf works", {
  expect_equal(
    sapply(tau, function(x) {
      left_inverse(cdf,
        at = x, low = -1, high = 7,
        tol = 1e-6, maxiter = 1000
      )
    }),
    ans
  )
  expect_equal(eval_quantile(d, at = tau), ans)
  expect_equal(eval_quantile(d, at = tau[permute]), ans[permute])
  expect_equal(eval_quantile(d, at = tau_left), ans_left)
  expect_equal(eval_quantile(d, at = tau_right), ans_right)
  expect_equal(
    eval_quantile(d, at = c(tau_right, tau_left, tau)),
    c(ans_right, ans_left, ans)
  )
  expect_equal(eval_quantile(d, at = 0.5), 3)
  expect_identical(eval_quantile(d, at = numeric(0L)), numeric(0L))
  expect_identical(eval_quantile(d, at = NA_real_), NA_real_)
  expect_identical(eval_quantile(d, at = NaN), NaN)
  expect_identical(
    eval_quantile(d, at = c(NA_real_, NA_real_)),
    c(NA_real_, NA_real_)
  )
})

d2 <- mix(dst_norm(0, 1), dst_norm(4, 1))

test_that("Quantile algorithm works with no discontinuities", {
  expect_equal(eval_quantile(d2, at = 0.5), 2)
  # eval_quantile(d2, at = 1:9/10)
  expect_equal(
    eval_quantile(d2, at = c(0, 0.5, NA, 1, 1.3)),
    c(-Inf, 2, NA, Inf, Inf)
  )
})