tests/testthat/test-univariate.R

In twolodzko/misster: Simple tools for dealing with missing values

context("Univariate")


test_that("The missing values are replaced.", {

  x <- c(NA, 1, 2, NA, 4, NA, NA, 7, NA)

  expect_false(anyNA(na_const(x)))
  expect_false(anyNA(na_mean(x)))
  expect_false(anyNA(na_median(x)))
  expect_message(expect_false(anyNA(na_mode(x))))
  expect_false(anyNA(na_interp(x)))

  expect_false(anyNA(na_interp(c(NA, 1, 2, NA))))
  expect_false(anyNA(na_interp(c(NA, NA, 1, 2, NA))))
  expect_false(anyNA(na_interp(c(NA, 1, 2, NA, NA))))
  expect_false(anyNA(na_interp(c(NA, 1, NA, 2, NA))))

  expect_false(anyNA(na_random(x, replace = TRUE)))
  expect_false(anyNA(na_random(c(1, 2, NA, NA), replace = FALSE)))
  expect_false(anyNA(na_density(x, preserve_var = TRUE)))
  expect_false(anyNA(na_density(x, preserve_var = FALSE)))

})


test_that("The missing values are replaced by the interpolating functions.", {

  x <- c(1, 2, NA, 4, NA, NA, 7)

  expect_false(anyNA(na_prev(x)))
  expect_false(anyNA(na_mid(x)))
  expect_warning(expect_false(anyNA(na_loess(x))))

})


test_that("Interpolation works as expected.", {

  x <- c(1, 2, NA, 4, NA, NA, 7)

  expect_equal(as.numeric(na_interp(x)), 1:7)
  expect_warning(expect_equal(as.numeric(na_loess(x)), 1:7, tolerance = 0.15))

})


test_that("C++ code identifies NAs correctly.", {

  test_cases <- list(
    c(NA, 1, 2, NA),
    c(NA, NA, 1, 2, NA),
    c(NA, 1, 2, NA, NA),
    c(NA, 1, NA, 2, NA),
    c(NA, NA, NA, 2, NA)
  )

  # na_interp

  expect_identical(na_positions(na_interp(test_cases[[1]])),
                   which(is.na(test_cases[[1]])))

  expect_identical(na_positions(na_interp(test_cases[[2]])),
                   which(is.na(test_cases[[2]])))

  expect_identical(na_positions(na_interp(test_cases[[3]])),
                   which(is.na(test_cases[[3]])))

  expect_identical(na_positions(na_interp(test_cases[[4]])),
                   which(is.na(test_cases[[4]])))

  expect_identical(na_positions(na_interp(test_cases[[5]])),
                   which(is.na(test_cases[[5]])))

  # na_mid

  expect_identical(na_positions(na_mid(test_cases[[1]])),
                   which(is.na(test_cases[[1]])))

  expect_identical(na_positions(na_mid(test_cases[[2]])),
                   which(is.na(test_cases[[2]])))

  expect_identical(na_positions(na_mid(test_cases[[3]])),
                   which(is.na(test_cases[[3]])))

  expect_identical(na_positions(na_mid(test_cases[[4]])),
                   which(is.na(test_cases[[4]])))

  expect_identical(na_positions(na_mid(test_cases[[5]])),
                   which(is.na(test_cases[[5]])))

  # na_prev

  expect_identical(na_positions(na_prev(test_cases[[1]])),
                   which(is.na(test_cases[[1]])))

  expect_identical(na_positions(na_prev(test_cases[[2]])),
                   which(is.na(test_cases[[2]])))

  expect_identical(na_positions(na_prev(test_cases[[3]])),
                   which(is.na(test_cases[[3]])))

  expect_identical(na_positions(na_prev(test_cases[[4]])),
                   which(is.na(test_cases[[4]])))

  expect_identical(na_positions(na_prev(test_cases[[5]])),
                   which(is.na(test_cases[[5]])))


  # no class attributes:

  expect_true(is_imputed(na_prev(c(1, 2, NA, 4, NA, NA, 7))))
  expect_true(is_imputed(na_mid(c(1, 2, NA, 4, NA, NA, 7))))
  expect_true(is_imputed(na_interp(c(1, 2, NA, 4, NA, NA, 7))))

})


test_that("After replacing missing values, vectors do not change their length.", {

  x <- c(NA, 1, 2, 2, NA, 4, 5, NA, NA, 7, 9, NA, NA)
  n <- length(x)

  expect_identical(length(na_const(x)), n)
  expect_identical(length(na_mean(x)), n)
  expect_identical(length(na_median(x)), n)
  expect_identical(length(na_mode(x)), n)
  expect_identical(length(na_interp(x)), n)
  expect_identical(length(na_mid(x)), n)
  expect_identical(length(na_prev(x)), n)
  expect_identical(length(na_random(x, replace = TRUE)), n)
  expect_identical(length(na_random(x, replace = FALSE)), n)
  expect_identical(length(na_density(x, preserve_var = TRUE)), n)
  expect_identical(length(na_density(x, preserve_var = FALSE)), n)

})