tests/testthat/test-uncategorized.R

In s3alfisc/fwildclusterboot: Fast Wild Cluster Bootstrap Inference for Linear Models

test_that("uncategorized tests", {
  skip_on_cran()
  skip_if_not(
    fwildclusterboot:::find_proglang("julia"),
    message = "skip test as julia installation not found."
  )

  # see this issue: https://github.com/s3alfisc/fwildclusterboot/issues/26

  set.seed(76903)
  dqrng::dqset.seed(98796)

  data1 <<-
    fwildclusterboot:::create_data(
      N = 500,
      N_G1 = 6,
      icc1 = 0.01,
      N_G2 = 4,
      icc2 = 0.01,
      numb_fe1 = 10,
      numb_fe2 = 10,
      seed = 9865
    )

  feols1 <-
    fixest::feols(
      proposition_vote ~ treatment + ideology1 + log_income + group_id2,
      data = data1,
      weights = ~weights
    )
  feols2 <-
    fixest::feols(
      proposition_vote ~ treatment + ideology1 + log_income + group_id2,
      data = data1,
      weights = data1$weights
    )

  feols3 <-
    fixest::feols(
      proposition_vote ~ treatment + ideology1 + log_income + group_id2,
      data = data1,
      cluster = ~group_id1
    )
  feols4 <-
    fixest::feols(
      proposition_vote ~ treatment + ideology1 + log_income + group_id2,
      data = data1,
      cluster = data1$group_id1
    )

  boot1 <-
    boottest(
      feols1,
      param = "treatment",
      B = 999,
      clustid = "group_id1"
    )
  boot2 <-
    boottest(
      feols2,
      param = "treatment",
      B = 999,
      clustid = "group_id1"
    )
  boot3 <-
    boottest(
      feols3,
      param = "treatment",
      B = 999,
      clustid = "group_id1"
    )
  boot4 <-
    boottest(
      feols4,
      param = "treatment",
      B = 999,
      clustid = "group_id1"
    )

  expect_equal(generics::tidy(boot1), generics::tidy(boot2))
  expect_equal(generics::tidy(boot3), generics::tidy(boot4))




  # test invariance of boottest() results to type of fixed effect variable
  # (numeric vs factor vs character)
  # test issue https://github.com/s3alfisc/fwildclusterboot/issues/14
  # raised by Timothée


  # Test 1: one cluster variable is numeric vs no cluster variable is numeric

  data(voters)

  to_char <- c("Q1_immigration", "Q2_defense", "group_id1")
  # sapply(voters[, to_char], class)

  voters_1 <<- voters
  voters_1$Q1_immigration <- as.numeric(voters_1$Q1_immigration)
  # sapply(voters_1[, to_char], class)

  feols_fit <-
    fixest::feols(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters)
  feols_fit_2 <-
    fixest::feols(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters_1)
  lfe_fit <-
    lfe::felm(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters)
  lfe_fit_2 <-
    lfe::felm(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters_1)

  boot1 <-
    suppressWarnings(boottest(
      feols_fit,
      clustid = c("Q1_immigration", "Q2_defense"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot2 <-
    suppressWarnings(boottest(
      feols_fit_2,
      clustid = c("Q1_immigration", "Q2_defense"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot3 <-
    suppressWarnings(boottest(
      feols_fit,
      clustid = c("Q1_immigration", "Q2_defense"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot4 <-
    suppressWarnings(boottest(
      feols_fit,
      clustid = c("Q1_immigration", "Q2_defense"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))

  expect_equal(boot1$t_stat, boot2$t_stat)
  expect_equal(boot2$t_stat, boot3$t_stat)
  expect_equal(boot3$t_stat, boot4$t_stat)
  expect_equal(boot4$t_stat, boot1$t_stat)


  # Test 2: one fixed effect is numeric vs no fixed effect is numeric

  data(voters)

  to_char <- c("Q1_immigration", "Q2_defense", "group_id1")
  # sapply(voters[, to_char], class)

  voters_1 <<- voters
  voters_1$Q2_defense <- as.numeric(voters_1$Q2_defense)
  # sapply(voters_1[, to_char], class)

  feols_fit <-
    fixest::feols(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters)
  feols_fit_2 <-
    fixest::feols(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters_1)
  lfe_fit <-
    lfe::felm(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters)
  lfe_fit_2 <-
    lfe::felm(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters_1)

  boot1 <-
    suppressWarnings(
      boottest(
        feols_fit,
        clustid = c("Q1_immigration"),
        B = 9999,
        param = "treatment",
        bootcluster = "min"
      )
    )
  boot2 <-
    suppressWarnings(
      boottest(
        feols_fit_2,
        clustid = c("Q1_immigration"),
        B = 9999,
        param = "treatment",
        bootcluster = "min"
      )
    )
  boot3 <-
    suppressWarnings(
      boottest(
        lfe_fit,
        clustid = c("Q1_immigration"),
        B = 9999,
        param = "treatment",
        bootcluster = "min"
      )
    )
  boot4 <-
    suppressWarnings(
      boottest(
        lfe_fit_2,
        clustid = c("Q1_immigration"),
        B = 9999,
        param = "treatment",
        bootcluster = "min"
      )
    )

  expect_equal(boot1$t_stat, boot2$t_stat)
  expect_equal(boot2$t_stat, boot3$t_stat)
  expect_equal(boot3$t_stat, boot4$t_stat)
  expect_equal(boot4$t_stat, boot1$t_stat)


  # Test 3: all fixed effects and cluster variables are numeric vs factors

  data(voters)

  to_char <- c("Q1_immigration", "Q2_defense", "group_id1")
  voters$group_id1 <- as.factor(voters$group_id1)
  # sapply(voters[, to_char], class)

  voters_1 <<- voters
  voters_1$Q1_immigration <- as.numeric(voters_1$Q1_immigration)
  voters_1$Q2_defense <- as.numeric(voters_1$Q2_defense)

  # sapply(voters_1[, to_char], class)

  feols_fit <-
    fixest::feols(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters)
  feols_fit_2 <-
    fixest::feols(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters_1)
  lfe_fit <-
    lfe::felm(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters)
  lfe_fit_2 <-
    lfe::felm(proposition_vote ~ treatment + log_income |
      Q2_defense, data = voters_1)

  boot1 <-
    suppressWarnings(boottest(
      feols_fit,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot2 <-
    suppressWarnings(boottest(
      feols_fit_2,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot3 <-
    suppressWarnings(boottest(
      lfe_fit,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot4 <-
    suppressWarnings(boottest(
      lfe_fit_2,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  expect_equal(boot1$t_stat, boot2$t_stat)
  expect_equal(boot2$t_stat, boot3$t_stat)
  expect_equal(boot3$t_stat, boot4$t_stat)
  expect_equal(boot4$t_stat, boot1$t_stat)


  # Test 4: Test 3, but now with two fixed effects


  data(voters)

  to_char <- c("Q1_immigration", "Q2_defense", "group_id1")
  voters$group_id1 <- as.factor(voters$group_id1)
  # sapply(voters[, to_char], class)

  voters_1 <<- voters
  voters_1$Q1_immigration <- as.numeric(voters_1$Q1_immigration)
  voters_1$Q2_defense <- as.numeric(voters_1$Q2_defense)

  # sapply(voters[, to_char], class)
  # sapply(voters_1[, to_char], class)

  feols_fit <-
    fixest::feols(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters
    )
  feols_fit_2 <-
    fixest::feols(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters_1
    )
  lfe_fit <-
    lfe::felm(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters
    )
  lfe_fit_2 <-
    lfe::felm(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters_1
    )

  boot1 <-
    suppressWarnings(boottest(
      feols_fit,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot2 <-
    suppressWarnings(boottest(
      feols_fit_2,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot3 <-
    suppressWarnings(boottest(
      lfe_fit,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  boot4 <-
    suppressWarnings(boottest(
      lfe_fit_2,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))

  expect_equal(boot1$t_stat, boot2$t_stat)
  expect_equal(boot2$t_stat, boot3$t_stat)
  expect_equal(boot3$t_stat, boot4$t_stat)
  expect_equal(boot4$t_stat, boot1$t_stat)



  # What if a fixed effect is a character?


  data(voters)

  to_char <- c("Q1_immigration", "Q2_defense", "group_id1")
  # sapply(voters[, to_char], class)

  voters_1 <<- voters
  voters_1$Q1_immigration <- as.character(voters_1$Q1_immigration)
  voters_1$Q2_defense <- as.character(voters_1$Q2_defense)

  # sapply(voters_1[, to_char], class)

  feols_fit <-
    fixest::feols(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters
    )
  feols_fit_2 <-
    fixest::feols(
      proposition_vote ~ treatment + log_income,
      fixef = c("Q1_immigration", "Q2_defense"),
      data = voters_1
    )
  lfe_fit <-
    lfe::felm(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters
    )
  lfe_fit_2 <-
    lfe::felm(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters_1
    )


  boot1 <-
    suppressWarnings(boottest(
      feols_fit,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))

  expect_error( # boot2 <-
    tmp <-
      suppressWarnings(
        boottest(
          feols_fit_2,
          clustid = c("Q1_immigration"),
          B = 9999,
          param = "treatment",
          bootcluster = "min"
        )
      )
  )
  boot3 <-
    suppressWarnings(boottest(
      lfe_fit,
      clustid = c("Q1_immigration"),
      B = 9999,
      param = "treatment",
      bootcluster = "min"
    ))
  expect_error( # boot4 <-
    suppressWarnings(
      boottest(
        lfe_fit_2,
        clustid = c("Q1_immigration"),
        B = 9999,
        param = "treatment",
        bootcluster = "min"
      )
    )
  )

  expect_equal(boot1$t_stat, boot3$t_stat)
  # expect_equal(boot2$t_stat, boot3$t_stat)
  # expect_equal(boot3$t_stat, boot4$t_stat)
  # expect_equal(boot4$t_stat, boot1$t_stat)


  # Test 4 with fe = ON in suppressWarnings(boottest()

  data(voters)

  to_char <- c("Q1_immigration", "Q2_defense", "group_id1")
  voters$group_id1 <- as.factor(voters$group_id1)
  # sapply(voters[, to_char], class)

  voters_1 <<- voters
  voters_1$Q1_immigration <- as.numeric(voters_1$Q1_immigration)
  voters_1$Q2_defense <- as.numeric(voters_1$Q2_defense)

  # sapply(voters_1[, to_char], class)

  feols_fit <-
    fixest::feols(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters
    )
  feols_fit_2 <-
    fixest::feols(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters_1
    )
  lfe_fit <-
    lfe::felm(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters
    )
  lfe_fit_2 <-
    lfe::felm(
      proposition_vote ~ treatment + log_income |
        Q1_immigration + Q2_defense,
      data = voters_1
    )

  boot1 <-
    suppressWarnings(
      boottest(
        feols_fit,
        clustid = c("Q1_immigration"),
        B = 9999,
        fe = "Q2_defense",
        param = "treatment",
        bootcluster = "min"
      )
    )
  boot2 <-
    suppressWarnings(
      boottest(
        feols_fit_2,
        clustid = c("Q1_immigration"),
        B = 9999,
        fe = "Q2_defense",
        param = "treatment",
        bootcluster = "min"
      )
    )
  boot3 <-
    suppressWarnings(
      boottest(
        lfe_fit,
        clustid = c("Q1_immigration"),
        B = 9999,
        fe = "Q2_defense",
        param = "treatment",
        bootcluster = "min"
      )
    )
  boot4 <-
    suppressWarnings(
      boottest(
        lfe_fit_2,
        clustid = c("Q1_immigration"),
        B = 9999,
        fe = "Q2_defense",
        param = "treatment",
        bootcluster = "min"
      )
    )

  expect_equal(boot1$t_stat, boot2$t_stat)
  expect_equal(boot2$t_stat, boot3$t_stat)
  expect_equal(boot3$t_stat, boot4$t_stat)
  expect_equal(boot4$t_stat, boot1$t_stat)

  # all NA cluster variables
  voters$group_id1 <- NA
  lm_fit <- lm(proposition_vote ~ treatment, data = voters)
  expect_error(boottest(
    lm_fit,
    B = 999,
    param = "treatment",
    clustid = "group_id1"
  ))
})

test_that("test vec2mat", {
  set.seed(5123)
  N <- 100
  x <- rnorm(N)
  cluster <- sample(letters[1:5], N, TRUE)
  g <- collapse::GRP(cluster, call = FALSE)
  mat1 <- fwildclusterboot:::vec2mat(x = x, group_id = g$group.id)

  mat2 <- aggregate(
    x = diag(x),
    by = list(g$group.id),
    FUN = "sum",
    simplify = TRUE
  )
  mat2 <- t(as.matrix(mat2))
  expect_equal(mat1, mat2[-1, ], ignore_attr = TRUE)
})