tests/testthat/test-resp_distributions.R

In resquin: Response Quality Indicators for Survey Research

testdata <- data.frame( # NA block and pattern block
  var_a = c(NA, 2, 3, 1,NA,NA, 5,2,1,4,1,3),
  var_b = c(NA,NA, 4,NA, 2,NA,NA,5,4,3,1,3),
  var_c = c(NA,NA,NA, 5,NA, 3,NA,3,5,2,1,4))

test_that("resp_distributions input tests", {
  expect_no_error(resp_distributions(testdata))
  expect_error(resp_distributions(testdata,min_valid_responses = 5),
               regexp = "must be between or equal to 0 and 1")
  expect_error(resp_distributions(testdata,min_valid_responses = -1),
               regexp = "must be between or equal to 0 and 1")
  expect_error(resp_distributions(testdata,min_valid_responses = "a"),
               regexp = "Argument 'min_valid_responses' must be numeric.")
  expect_error(resp_distributions(testdata,min_valid_responses = T),
               regexp = "Argument 'min_valid_responses' must be numeric.")
  expect_error(resp_distributions(data.frame(var_a = c(NA,NA),
                                  var_b = c(NA,NA))),
               regexp = "No response quality indicators were calculated as the proportion of")
  expect_error(resp_distributions(as.matrix(testdata)),
               regexp = "x must be a data.frame or a tibble")
  expect_error(resp_distributions(T),
               regexp = "x must be a data.frame or a tibble")
  expect_error(resp_distributions(data.frame(var_a = c(1.5,2),
                                  var_b = c(2,3))),
               regexp = "Non-integer data found in following columns")
  expect_error(resp_distributions(data.frame(var_a = c("a","b"),
                                  var_b = c(2,3))),
               regexp = "Non-integer data found in following columns")
  expect_error(resp_distributions(data.frame(var_a = c(T,F),
                                  var_b = c(2,3))),
               regexp = "Non-integer data found in following columns")
  # Extra tests for id input
  expect_error(resp_distributions(x = testdata,
                                  id = 0),
               regexp = "id is not of type logical with length one or a numeric or character vector with length equal to the number of rows of x.")
  expect_error(resp_distributions(x = testdata,
                                  id = c(T,T)),
               regexp = "id is not of type numeric or character")
  expect_error(resp_distributions(x = testdata,
                                  id = c(1,2)),
               regexp = "Supply an `id` variable with the same number of elements as there are rows in x.")
  expect_error(resp_distributions(x = testdata,
                                  id = c(1:11,11)),
               regexp = "Supply an `id` variable which uniquely identifies each respondent by position.")
})

test_that("resp_distributions output tests", {
  expect_equal(resp_distributions(testdata,
                       min_valid_responses = 0)$n_na,
               c(3,2,1,1,2,2,2,0,0,0,0,0))
  expect_equal(resp_distributions(testdata,
                       min_valid_responses = 1)$prop_na,
               c(1,2/3,1/3,1/3,2/3,2/3,2/3,0,0,0,0,0))
  expect_equal(resp_distributions(testdata,
                       min_valid_responses = 0)$ii_mean,
               c(NA,2,(3+4)/2,(1+5)/2,2,3,5,
                 (2+5+3)/3,
                 (1+4+5)/3,
                 (4+3+2)/3,
                 (1+1+1)/3,
                 (3+3+4)/3))
  expect_equal(resp_distributions(testdata,
                       min_valid_responses = 0)$ii_median,
               c(NA,2,3.5,(1+5)/2,2,3,5,
                 median(c(2,5,3)),
                 median(c(1,4,5)),
                 median(c(4,3,2)),
                 median(c(1,1,1)),
                 median(c(3,3,4))))
  expect_equal(resp_distributions(testdata,
               min_valid_responses = 0)$ii_sd,
               testdata |>
                 apply(1,sd,na.rm=T))
  expect_equal(resp_distributions(testdata,
                       min_valid_responses = 0)$mahal,
    {mahal_res <- resquin:::mahalanobis_na(x = testdata,
                           center = colMeans(testdata,na.rm=T),
                           cov = cov(testdata,
                                     use = "pairwise.complete.obs"))
    mahal_res[1] <- NA
    mahal_res})
  expect_equal(object = resp_distributions(testdata,min_valid_responses = 1)$mahal |> stats::na.omit() |> as.numeric(),
               expected = testdata[8:12,] |> stats::mahalanobis(colMeans(testdata[8:12,] ),cov(testdata[8:12,] )) |> sqrt() |> as.numeric())
})


# Test for large data sets
#testdata_large <- purrr::map(1:50,
#                             \(cur_var){
#                               sample.int(5,size = 500000,replace = T)
#                             }) |>
#  as.data.frame()
#
#
#test_that("resp_distributions_output_test_large",{
#  skip_on_cran()
#  expect_no_error(resp_distributions(testdata_large))
#})