tests/testthat/test-samplingweights.r

In sdcTable: Methods for Statistical Disclosure Control in Tabular Data

context("test problem generation with sampling weights")

test_that("sdcprobs with sampling weights are correctly generated", {
  skip_on_cran()

  set.seed(10)
  utils::data("microdata1", package = "sdcTable")
  microdata1$sampweights <- sample(1:10, nrow(microdata1), replace = TRUE)

  dims <- list(
    region = sdcHierarchies::hier_create(root = "Total", nodes = LETTERS[1:4]),
    gender = sdcHierarchies::hier_create(root = "Total", nodes = c("male", "female")))

  p <- makeProblem(
    data = microdata1,
    dimList = dims,
    numVarInd = "val",
    sampWeightInd = "sampweights")

  expect_is(p, "sdcProblem")
  expect_equal(get.problemInstance(p@problemInstance, "nrVars"), 15)

  df <- sdcProb2df(p, addDups = TRUE, addNumVars = TRUE, dimCodes = "original")
  expect_equal(df$freq[1], sum(microdata1$sampweights))
  expect_equal(df$val[1], sum(microdata1$sampweights * microdata1$val))

  # starting from a complete table
  df_full <- df[, c("region", "gender", "freq", "val")]
  p <- makeProblem(
    data = df_full,
    dimList = dims,
    numVarInd = "val",
    freqVarInd = "freq")
  expect_is(p, "sdcProblem")
  expect_equal(get.problemInstance(p@problemInstance, "nrVars"), 15)
  df <- sdcProb2df(p, addDups = TRUE, addNumVars = TRUE, dimCodes = "original")
  expect_equal(df$freq[1], sum(microdata1$sampweights))
  expect_equal(df$val[1], sum(microdata1$sampweights * microdata1$val))

  # check that sampling weights are ignored in this case
  df_full$sampweights <- sample(1:10, nrow(df_full), replace = TRUE)
  p <- makeProblem(
    data = df_full,
    dimList = dims,
    numVarInd = "val",
    freqVarInd = "freq",
    sampWeightInd = "sampweights")
  expect_is(p, "sdcProblem")
  expect_equal(get.problemInstance(p@problemInstance, "nrVars"), 15)
  df <- sdcProb2df(p, addDups = TRUE, addNumVars = TRUE, dimCodes = "original")
  expect_equal(df$freq[1], sum(microdata1$sampweights))
  expect_equal(df$val[1], sum(microdata1$sampweights * microdata1$val))
})