tests/testthat/test-GammaDistribution.R

In rdecision: Decision Analytic Modelling in Health Economics

test_that("illegal initializations are rejected", {
  k <- 9.0
  theta <- 0.5
  expect_silent(GammaDistribution$new(k, theta))
  expect_error(GammaDistribution$new("9", theta), class = "shape_not_numeric")
  expect_error(GammaDistribution$new(k, "0.5"), class = "scale_not_numeric")
  expect_error(
    GammaDistribution$new(-1.0, theta), class = "shape_not_supported"
  )
  expect_error(GammaDistribution$new(k, 0.0), class = "scale_not_supported")
})

test_that("distribution name is correct", {
  k <- 9.0
  theta <- 0.5
  g <- GammaDistribution$new(k, theta)
  expect_identical(g$distribution(), "Ga(9,0.5)")
})

test_that("mean, mode, sd and quantiles are returned correctly", {
  k <- 9.0
  theta <- 0.5
  g <- GammaDistribution$new(k, theta)
  expect_intol(g$mean(), k * theta, 0.01)
  expect_intol(g$SD(), sqrt(k) * theta, 0.01)
  expect_intol(g$mode(), (k - 1.0) * theta, 0.01)
  probs <- c(0.025, 0.975)
  q <- g$quantile(probs)
  expect_intol(q[[1L]], 2.06, 0.01)
  expect_intol(q[[2L]], 7.88, 0.01)
})

test_that("quantile function checks inputs and has correct output", {
  k <- 9.0
  theta <- 0.5
  g <- GammaDistribution$new(k, theta)
  probs <- c(0.1, 0.2, 0.5)
  expect_silent(g$quantile(probs))
  probs <- c(0.1, NA, 0.5)
  expect_error(g$quantile(probs), class = "probs_not_defined")
  probs <- c(0.1, "boo", 0.5)
  expect_error(g$quantile(probs), class = "probs_not_numeric")
  probs <- c(0.1, 0.4, 1.5)
  expect_error(g$quantile(probs), class = "probs_out_of_range")
  probs <- c(0.1, 0.2, 0.5)
  expect_length(g$quantile(probs), 3L)
})

test_that("random sampling is from a Gamma distribution", {
  k <- 9.0
  theta <- 0.5
  n <- 1000L
  g <- GammaDistribution$new(k, theta)
  # mean
  g$sample(TRUE)
  expect_equal(g$r(), 4.5)
  # random sampling
  samp <- vapply(seq_len(n), FUN.VALUE = 1.0, FUN = function(i) {
    g$sample()
    rv <- g$r()
    return(rv)
  })
  expect_length(samp, n)
  # 99.9% confidence limits; expected test failure rate is 0.1%;
  # skip for CRAN
  skip_on_cran()
  ht <- ks.test(samp, stats::rgamma(n, shape = k, scale = theta))
  expect_gt(ht$p.value, 0.001)
})