tests/testthat/test-pls.R

In recipes: Preprocessing and Feature Engineering Steps for Modeling

library(testthat)
library(recipes)

skip_if_not_installed("modeldata")

## -----------------------------------------------------------------------------

data(biomass, package = "modeldata")

biom_tr <- biomass %>%
  dplyr::filter(dataset == "Training") %>%
  dplyr::select(-dataset, -sample)
biom_te <- biomass %>%
  dplyr::filter(dataset == "Testing") %>%
  dplyr::select(-dataset, -sample, -HHV)

data(cells, package = "modeldata")

cell_tr <- cells %>%
  dplyr::filter(case == "Train") %>%
  dplyr::select(-case)
cell_te <- cells %>%
  dplyr::filter(case == "Test") %>%
  dplyr::select(-case, -class)

load(test_path("test_pls_new.RData"))


## -----------------------------------------------------------------------------

test_that("PLS, dense loadings", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(HHV ~ ., data = biom_tr) %>%
    step_pls(all_predictors(), outcome = "HHV", num_comp = 3)

  rec <- prep(rec)

  expect_equal(
    names(rec$steps[[1]]$res),
    c("mu", "sd", "coefs", "col_norms")
  )

  tr_new <- bake(rec, new_data = NULL, all_predictors())
  expect_equal(tr_new, bm_pls_tr)
  te_new <- bake(rec, biom_te)
  expect_equal(te_new, bm_pls_te)
})


test_that("PLS, dense loadings, multiple outcomes", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(HHV + carbon ~ ., data = biom_tr) %>%
    step_pls(all_predictors(), outcome = c("HHV", "carbon"), num_comp = 3)

  rec <- prep(rec)

  expect_equal(
    names(rec$steps[[1]]$res),
    c("mu", "sd", "coefs", "col_norms")
  )

  tr_new <- bake(rec, new_data = NULL, all_predictors())
  expect_equal(tr_new, bm_pls_multi_tr)
  te_new <- bake(rec, biom_te %>% select(-carbon))
  expect_equal(te_new, bm_pls_multi_te)
})


test_that("PLS, sparse loadings", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(HHV ~ ., data = biom_tr) %>%
    step_pls(all_predictors(), outcome = "HHV", num_comp = 3,
             predictor_prop = 3 / 5)

  rec <- prep(rec)

  expect_equal(
    names(rec$steps[[1]]$res),
    c("mu", "sd", "coefs", "col_norms")
  )

  tr_new <- bake(rec, new_data = NULL, all_predictors())
  expect_equal(tr_new, bm_spls_tr)
  te_new <- bake(rec, biom_te)
  expect_equal(te_new, bm_spls_te)
})


test_that("PLS, dense loadings, multiple outcomes", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(HHV + carbon ~ ., data = biom_tr) %>%
    step_pls(all_predictors(), outcome = c("HHV", "carbon"), num_comp = 3,
             predictor_prop = 3 / 5)

  rec <- prep(rec)

  expect_equal(
    names(rec$steps[[1]]$res),
    c("mu", "sd", "coefs", "col_norms")
  )

  tr_new <- bake(rec, new_data = NULL, all_predictors())
  expect_equal(tr_new, bm_spls_multi_tr)
  te_new <- bake(rec, biom_te %>% select(-carbon))
  expect_equal(te_new, bm_spls_multi_te)
})

## -----------------------------------------------------------------------------

test_that("PLS-DA, dense loadings", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(class ~ ., data = cell_tr) %>%
    step_pls(all_predictors(), outcome = "class", num_comp = 3)

  rec <- prep(rec)

  expect_equal(
    names(rec$steps[[1]]$res),
    c("mu", "sd", "coefs", "col_norms")
  )

  tr_new <- bake(rec, new_data = NULL, all_predictors())
  expect_equal(tr_new, cell_plsda_tr)
  te_new <- bake(rec, cell_te)
  expect_equal(te_new, cell_plsda_te)
})


test_that("PLS-DA, dense loadings, multiple outcomes", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(class + case ~ ., data = cells) %>%
    step_pls(all_predictors(), outcome = c("class", "case"), num_comp = 3)

  expect_snapshot(error = TRUE, prep(rec))
})


test_that("PLS-DA, sparse loadings", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(class ~ ., data = cell_tr) %>%
    step_pls(all_predictors(), outcome = "class", num_comp = 3,
             predictor_prop = 50 / 56)

  rec <- prep(rec)

  expect_equal(
    names(rec$steps[[1]]$res),
    c("mu", "sd", "coefs", "col_norms")
  )

  tr_new <- bake(rec, new_data = NULL, all_predictors())
  expect_equal(tr_new, cell_splsda_tr)
  te_new <- bake(rec, cell_te)
  expect_equal(te_new, cell_splsda_te)
})


test_that("PLS-DA, sparse loadings, multiple outcomes", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(class + case ~ ., data = cells) %>%
    step_pls(all_predictors(), outcome = c("class", "case"), num_comp = 3,
             predictor_prop = 50 / 56)

  expect_snapshot(error = TRUE, prep(rec))
})

## -----------------------------------------------------------------------------

test_that("No PLS", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(class ~ ., data = cell_tr) %>%
    step_pls(all_predictors(), outcome = "class", num_comp = 0)

  rec <- prep(rec)

  expect_null(
    rec$steps[[1]]$res
  )
  pred_names <- summary(rec)$variable[summary(rec)$role == "predictor"]

  tr_new <- bake(rec, new_data = NULL, all_predictors())
  expect_equal(names(tr_new), pred_names)
  te_new <- bake(rec, cell_te, all_predictors())
  expect_equal(names(te_new), pred_names)
})

## -----------------------------------------------------------------------------

test_that("tidy method", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(HHV ~ ., data = biom_tr) %>%
    step_pls(all_predictors(), outcome = "HHV", num_comp = 3, id = "dork")

  tidy_pre <- tidy(rec, number = 1)
  exp_pre <- tibble::tribble(
    ~terms, ~value, ~component, ~id,
    "all_predictors()", NA_real_, NA_character_, "dork"
  )
  expect_equal(tidy_pre, exp_pre)

  rec <- prep(rec)
  tidy_post <- tidy(rec, number = 1)
  exp_post <-
    tibble::tribble(
      ~terms, ~value, ~component, ~id,
      "carbon", 0.82813459059393, "PLS1", "dork",
      "carbon", 0.718469477422311, "PLS2", "dork",
      "carbon", 0.476111929729498, "PLS3", "dork",
      "hydrogen", -0.206963356355556, "PLS1", "dork",
      "hydrogen", 0.642998926998282, "PLS2", "dork",
      "hydrogen", 0.262836631090453, "PLS3", "dork",
      "oxygen", -0.49241242430895, "PLS1", "dork",
      "oxygen", 0.299176769170812, "PLS2", "dork",
      "oxygen", 0.418081563632953, "PLS3", "dork",
      "nitrogen", -0.122633995804743, "PLS1", "dork",
      "nitrogen", -0.172719084680244, "PLS2", "dork",
      "nitrogen", 0.642403301090588, "PLS3", "dork",
      "sulfur", 0.11768677260853, "PLS1", "dork",
      "sulfur", -0.217341766567037, "PLS2", "dork",
      "sulfur", 0.521114256955661, "PLS3", "dork"
    )
  expect_equal(tidy_post, exp_post, tolerance = 0.01)
})

test_that("check_name() is used", {
  skip_if_not_installed("mixOmics")
  dat <- mtcars
  dat$PLS1 <- dat$mpg

  rec <- recipe(~ ., data = dat) %>%
    step_pls(mpg, disp, vs, outcome = "am")

  expect_snapshot(
    error = TRUE,
    prep(rec, training = dat)
  )
})

## -----------------------------------------------------------------------------

test_that("Deprecation warning", {
  expect_snapshot(error = TRUE,
    recipe(~ ., data = mtcars) %>%
      step_pls(outcome = "mpg", preserve = TRUE)
  )
})

test_that("tunable", {
  rec <-
    recipe(Species ~ ., data = iris) %>%
    step_pls(all_predictors(), outcome = "Species")
  rec_param <- tunable.step_pls(rec$steps[[1]])
  expect_equal(rec_param$name, c("num_comp", "predictor_prop"))
  expect_true(all(rec_param$source == "recipe"))
  expect_true(is.list(rec_param$call_info))
  expect_equal(nrow(rec_param), 2)
  expect_equal(
    names(rec_param),
    c("name", "call_info", "source", "component", "component_id")
  )
})

test_that("Do nothing for num_comps = 0 and keep_original_cols = FALSE (#1152)", {
  rec <- recipe(carb ~ ., data = mtcars) %>%
    step_pls(all_predictors(), outcome = "carb",
             num_comp = 0, keep_original_cols = FALSE) %>%
    prep()

  res <- bake(rec, new_data = NULL)

  expect_identical(res, tibble::as_tibble(mtcars))
})

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

test_that("bake method errors when needed non-standard role columns are missing", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(HHV ~ ., data = biom_tr) %>%
    step_pls(carbon, outcome = "HHV", num_comp = 3) %>%
    update_role(carbon, new_role = "potato") %>%
    update_role_requirements(role = "potato", bake = FALSE)

  rec <- prep(rec)

  expect_error(bake(rec, new_data = biom_tr[, c(-1)]),
               class = "new_data_missing_column")
})

test_that("empty printing", {
  rec <- recipe(mpg ~ ., mtcars)
  rec <- step_pls(rec, outcome = "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_pls(rec1, outcome = "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_pls(rec, outcome = "mpg")

  expect <- tibble(
    terms = character(),
    value = double(),
    component = character(),
    id = character()
  )

  expect_identical(tidy(rec, number = 1), expect)

  rec <- prep(rec, mtcars)

  expect_identical(tidy(rec, number = 1), expect)
})

test_that("keep_original_cols works", {
  skip_if_not_installed("mixOmics")
  new_names <- c("vs", "PLS1")

  rec <- recipe(vs ~ mpg, mtcars) %>%
    step_pls(all_predictors(), outcome = "vs", keep_original_cols = FALSE)

  rec <- prep(rec)
  res <- bake(rec, new_data = NULL)

  expect_equal(
    colnames(res),
    new_names
  )

  rec <- recipe(vs ~ mpg, mtcars) %>%
    step_pls(all_predictors(), outcome = "vs", keep_original_cols = TRUE)

  rec <- prep(rec)
  res <- bake(rec, new_data = NULL)

  expect_equal(
    colnames(res),
    c("mpg", new_names)
  )
})

test_that("keep_original_cols - can prep recipes with it missing", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(vs ~ mpg, mtcars) %>%
    step_pls(all_predictors(), outcome = "vs")

  rec$steps[[1]]$keep_original_cols <- NULL

  expect_snapshot(
    rec <- prep(rec)
  )

  expect_error(
    bake(rec, new_data = mtcars),
    NA
  )
})

test_that("printing", {
  skip_if_not_installed("mixOmics")
  rec <- recipe(HHV ~ ., data = biom_tr) %>%
    step_pls(all_predictors(), outcome = "HHV", num_comp = 3)

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

test_that("tunable is setup to work with extract_parameter_set_dials", {
  skip_if_not_installed("dials")
  rec <- recipe(mpg ~ ., data = mtcars) %>%
    step_pls(
      all_predictors(),
      outcome = "mpg",
      num_comp = hardhat::tune(), predictor_prop = hardhat::tune()
    )

  params <- extract_parameter_set_dials(rec)

  expect_s3_class(params, "parameters")
  expect_identical(nrow(params), 2L)
})