tests/testthat/test-lencode_glm.R

In topepo/embed: Extra Recipes for Encoding Predictors

source(testthat::test_path("make_example_data.R"))
source(testthat::test_path("test-helpers.R"))

test_that("factor encoded predictor", {
  class_test <- recipe(x2 ~ ., data = ex_dat) %>%
    step_lencode_glm(x3, outcome = vars(x2), id = "id") %>%
    prep(training = ex_dat, retain = TRUE)
  tr_values <- bake(class_test, new_data = NULL)$x3
  new_values <- bake(class_test, new_data = new_dat)
  expect_snapshot(
    new_values_ch <- bake(class_test, new_data = new_dat_ch)
  )
  key <- class_test$steps[[1]]$mapping
  td_obj <- tidy(class_test, number = 1)

  expect_equal("x3", names(key))

  expect_equal(
    length(unique(ex_dat$x3)) + 1,
    nrow(key$x3)
  )
  expect_true(sum(key$x3$..level == "..new") == 1)

  expect_true(is.numeric(tr_values))

  expect_equal(
    new_values$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_ch$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_ch$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values_ch$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )

  expect_equal(
    td_obj$level,
    key$x3$..level
  )
  expect_equal(
    td_obj$value,
    key$x3$..value
  )
})

test_that("character encoded predictor", {
  class_test <- recipe(x2 ~ ., data = ex_dat_ch) %>%
    step_lencode_glm(x3, outcome = vars(x2)) %>%
    prep(training = ex_dat_ch, retain = TRUE)
  tr_values <- bake(class_test, new_data = NULL)$x3
  expect_snapshot(
    new_values <- bake(class_test, new_data = new_dat_ch)
  )
  new_values_fc <- bake(class_test, new_data = new_dat)
  key <- class_test$steps[[1]]$mapping
  td_obj <- tidy(class_test, number = 1)

  expect_equal("x3", names(key))

  expect_equal(
    length(unique(ex_dat$x3)) + 1,
    nrow(key$x3)
  )
  expect_true(sum(key$x3$..level == "..new") == 1)

  expect_true(is.numeric(tr_values))

  expect_equal(
    new_values$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_fc$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_fc$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values_fc$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )

  expect_equal(
    td_obj$level,
    key$x3$..level
  )
  expect_equal(
    td_obj$value,
    key$x3$..value
  )
})

test_that("factor encoded predictor", {
  reg_test <- recipe(x1 ~ ., data = ex_dat) %>%
    step_lencode_glm(x3, outcome = vars(x1)) %>%
    prep(training = ex_dat, retain = TRUE)
  tr_values <- bake(reg_test, new_data = NULL)$x3
  new_values <- bake(reg_test, new_data = new_dat)
  expect_snapshot(
    new_values_ch <- bake(reg_test, new_data = new_dat_ch)
  )
  td_obj <- tidy(reg_test, number = 1)

  key <- reg_test$steps[[1]]$mapping

  expect_equal("x3", names(key))

  expect_equal(
    length(unique(ex_dat$x3)) + 1,
    nrow(key$x3)
  )
  expect_true(sum(key$x3$..level == "..new") == 1)

  expect_true(is.numeric(tr_values))

  expect_equal(
    new_values$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_ch$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_ch$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values_ch$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )

  expect_equal(
    td_obj$level,
    key$x3$..level
  )
  expect_equal(
    td_obj$value,
    key$x3$..value
  )
})

test_that("character encoded predictor", {
  reg_test <- recipe(x1 ~ ., data = ex_dat_ch) %>%
    step_lencode_glm(x3, outcome = vars(x1)) %>%
    prep(training = ex_dat_ch, retain = TRUE)
  tr_values <- bake(reg_test, new_data = NULL)$x3
  new_values <- bake(reg_test, new_data = new_dat_ch)
  new_values_fc <- bake(reg_test, new_data = new_dat)
  key <- reg_test$steps[[1]]$mapping
  td_obj <- tidy(reg_test, number = 1)

  expect_equal("x3", names(key))

  expect_equal(
    length(unique(ex_dat$x3)) + 1,
    nrow(key$x3)
  )
  expect_true(sum(key$x3$..level == "..new") == 1)

  expect_true(is.numeric(tr_values))

  expect_equal(
    new_values$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_fc$x3[1],
    key$x3$..value[key$x3$..level == "..new"]
  )
  expect_equal(
    new_values_fc$x3[2],
    key$x3$..value[key$x3$..level == levels(ex_dat$x3)[1]]
  )
  expect_equal(
    new_values_fc$x3[3],
    key$x3$..value[key$x3$..level == "..new"]
  )

  expect_equal(
    td_obj$level,
    key$x3$..level
  )
  expect_equal(
    td_obj$value,
    key$x3$..value
  )
})

test_that("bad args", {
  three_class <- iris
  three_class$fac <- rep(letters[1:3], 50)
  three_class$logical <- rep(c(TRUE, FALSE), 75)

  expect_snapshot(
    error = TRUE,
    recipe(Species ~ ., data = three_class) %>%
      step_lencode_glm(Sepal.Length, outcome = vars(Species)) %>%
      prep(training = three_class, retain = TRUE)
  )
})

test_that("case weights", {
  wts_int <- rep(c(0, 1), times = c(100, 400))

  ex_dat_cw <- ex_dat %>%
    mutate(wts = importance_weights(wts_int))

  class_test <- recipe(x2 ~ ., data = ex_dat_cw) %>%
    step_lencode_glm(x3, outcome = vars(x2), id = "id") %>%
    prep(training = ex_dat_cw, retain = TRUE)

  ref_mod <- glm(
    x2 ~ 0 + x3,
    data = ex_dat_cw,
    family = binomial,
    na.action = na.omit, weights = ex_dat_cw$wts
  )

  expect_equal(
    -unname(coef(ref_mod)),
    slice_head(class_test$steps[[1]]$mapping$x3, n = -1)$..value
  )

  expect_snapshot(class_test)
})

# Infrastructure ---------------------------------------------------------------

test_that("bake method errors when needed non-standard role columns are missing", {
  rec <- recipe(x2 ~ ., data = ex_dat) %>%
    step_lencode_glm(x3, outcome = vars(x2)) %>%
    update_role(x3, new_role = "potato") %>%
    update_role_requirements(role = "potato", bake = FALSE)
  
  rec_trained <- prep(rec, training = ex_dat, verbose = FALSE)
  
  expect_error(
    bake(rec_trained, new_data = ex_dat[, -3]),
    class = "new_data_missing_column"
  )
})

test_that("empty printing", {
  rec <- recipe(mpg ~ ., mtcars)
  rec <- step_lencode_glm(rec, outcome = vars(mpg))
  
  expect_snapshot(rec)
  
  rec <- prep(rec, mtcars)
  
  expect_snapshot(rec)
})

test_that("empty selection prep/bake is a no-op", {
  rec1 <- recipe(mpg ~ ., mtcars)
  rec2 <- step_lencode_glm(rec1, outcome = vars(mpg))
  
  rec1 <- prep(rec1, mtcars)
  rec2 <- prep(rec2, mtcars)
  
  baked1 <- bake(rec1, mtcars)
  baked2 <- bake(rec2, mtcars)
  
  expect_identical(baked1, baked2)
})

test_that("empty selection tidy method works", {
  rec <- recipe(mpg ~ ., mtcars)
  rec <- step_lencode_glm(rec, outcome = vars(mpg))
  
  expect <- tibble(
    terms = character(),
    level = character(),
    value = double(),
    id = character()
  )
  
  expect_identical(tidy(rec, number = 1), expect)
  
  rec <- prep(rec, mtcars)
  
  expect_identical(tidy(rec, number = 1), expect)
})

test_that("printing", {
  rec <- recipe(x2 ~ ., data = ex_dat_ch) %>%
    step_lencode_glm(x3, outcome = vars(x2))

  expect_snapshot(print(rec))
  expect_snapshot(prep(rec))
})