tests/testthat/test-distribution_normal.R

In dynverse/dynparam: Creating Meta-Information for Parameters

context("test-distribution_normal")

test_that("normal_distribution works with fixed values", {
  dis <- normal_distribution(mean = 50, sd = 10)
  expect_match(as.character(dis), "N\\(50[\\.0]*, 10[\\.0]*\\)")

  dis <- normal_distribution(mean = 5, sd = 1, lower = -Inf, upper = Inf)
  dis <- normal_distribution(mean = 5, sd = 1, lower = 0, upper = Inf)
  dis <- normal_distribution(mean = 5, sd = 1, lower = -Inf, upper = 0)
})

test_that("normal_distribution works with random values", {
  mean <- runif(1, 50, 100)
  sd <- runif(1, 5, 10)

  # create distribution
  dis <- normal_distribution(mean = mean, sd = sd)
  expect_lte(abs(dis$mean - mean), 1e-5)
  expect_lte(abs(dis$sd - sd), 1e-5)

  # transform to list
  lis <- as.list(dis)
  expect_equal(lis$type, gsub("_distribution", "", class(dis))[[1]])
  expect_lte(abs(lis$mean - mean), 1e-5)
  expect_lte(abs(lis$sd - sd), 1e-5)

  # transform back to distribution
  dis2 <- as_distribution(lis)
  expect_equal(class(dis2), class(dis))
  expect_lte(abs(dis2$mean - mean), 1e-5)
  expect_lte(abs(dis2$sd - sd), 1e-5)

  # test dist and quan functions
  dfun <- distribution_function(dis)
  qfun <- quantile_function(dis)

  expect_equal(qfun(0), -Inf)
  expect_lte(abs(qfun(0.5) - mean), 1e-5)
  expect_equal(qfun(1), Inf)
  expect_true(all(abs(dfun(c(-Inf, mean, Inf)) - c(0, .5, 1)) < 1e-5))
})

test_that("normal_distribution with limits works with random values", {
  mean <- runif(1, 50, 100)
  sd <- runif(1, 5, 10)
  lower <- runif(1, 0, 49)
  upper <- runif(1, 101, 200)

  # create distribution
  dis <- normal_distribution(mean = mean, sd = sd, lower = lower, upper = upper)
  expect_lte(abs(dis$lower - lower), 1e-5)
  expect_lte(abs(dis$upper - upper), 1e-5)

  # transform to list
  lis <- as.list(dis)
  expect_equal(lis$type, gsub("_distribution", "", class(dis))[[1]])
  expect_lte(abs(lis$lower - lower), 1e-5)
  expect_lte(abs(lis$upper - upper), 1e-5)

  # transform back to distribution
  dis2 <- as_distribution(lis)
  expect_equal(class(dis2), class(dis))
  expect_lte(abs(dis2$lower - lower), 1e-5)
  expect_lte(abs(dis2$upper - upper), 1e-5)
})

test_that("normal_distribution errors when expected", {
  expect_error(normal_distribution(mean = -Inf, sd = 0), "mean is not bounded by \\(-Inf,Inf\\)")
  expect_error(normal_distribution(mean = NA, sd = NA), "is not a single numeric value")
  expect_error(normal_distribution(mean = 0, sd = Inf), "sd is not bounded by \\(-Inf,Inf\\)")

  expect_error(normal_distribution(mean = 1, sd = 1, lower = Inf, upper = 0), "lower is not bounded by \\[-Inf,Inf\\)")
  expect_error(normal_distribution(mean = 1, sd = 1, lower = NA, upper = NA), "is not a single numeric value")
  expect_error(normal_distribution(mean = 1, sd = 1, lower = 0, upper = -Inf), "upper is not bounded by \\(-Inf,Inf\\]")
  expect_error(normal_distribution(mean = 1, sd = 1, lower = 10, upper = 0), "lower not less than upper")
})