tests/testthat/test-familiar_data_creation.R

In familiar: End-to-End Automated Machine Learning and Model Evaluation

# Don't perform any further tests on CRAN due to time of running the complete test.
testthat::skip_on_cran()
testthat::skip_on_ci()

familiar_data_creation_unit_test <- function(outcome_type) {
  if (outcome_type == "count") {
    learner <- "glm_poisson"
  } else if (outcome_type == "continuous") {
    learner <- "glm_gaussian"
  } else if (outcome_type == "binomial") {
    learner <- "glm_logistic"
  } else if (outcome_type == "multinomial") {
    learner <- "glm_multinomial"
  } else if (outcome_type == "survival") {
    learner <- "cox"
  } else {
    ..error_outcome_type_not_implemented(outcome_type)
  }

  # Create a dataset using the good dataset.
  data <- familiar:::test_create_good_data(outcome_type)

  # Train a simple linear GLM using the good dataset.
  fam_model <- suppressWarnings(familiar:::test_train(
    data = data,
    cluster_method = "none",
    imputation_method = "simple",
    hyperparameter_list = list("sign_size" = familiar:::get_n_features(data)),
    learner = learner,
    time_max = 1832))
  
  testthat::test_that(
    paste0(
      "1. familiarData for the ", outcome_type, " outcome can be created from ",
      "good data and the model created using the complete data set."),
    {
      # Test with small, but good dataset.
      fam_data <- suppressWarnings(familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_small_good_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE))
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )
  
  testthat::test_that(
    paste0(
      "2. familiarData for the ", outcome_type, " outcome can be created from ",
      "single-sample data and the model created using the complete data set."), 
    {
      # Test with one-sample dataset.
      fam_data <- familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_one_sample_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE
      )
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )

  testthat::test_that(
    paste0(
      "3. familiarData for the ", outcome_type, " outcome can be created from ",
      "empty data and the model created using the complete data set."), 
    {
    # Test with empty dataset
      fam_data <- familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_empty_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE)
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )

  testthat::test_that(
    paste0(
      "4. familiarData for the ", outcome_type, " outcome can be created from ",
      "the bad dataset and the model created using the complete data set."), 
    {
      # Test with bad dataset
      fam_data <- suppressWarnings(familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_small_bad_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE))
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )
  
  # Create a dataset using the one-feature dataset.
  data <- familiar:::test_create_single_feature_data(outcome_type = outcome_type)
  
  # Train a simple linear GLM using the one feature dataset.
  fam_model <- suppressWarnings(familiar:::test_train(
    data = data,
    cluster_method = "none",
    imputation_method = "simple",
    hyperparameter_list = list("sign_size" = familiar:::get_n_features(data)),
    learner = learner,
    time_max = 1832))

  testthat::test_that(
    paste0(
      "5. familiarData for the ", outcome_type, " outcome can be created from ",
      "the one-feature dataset and the model created using the same data set."),
    {
      # Test with one-feature dataset.
      fam_data <- suppressWarnings(familiar::as_familiar_data(
        object = fam_model,
        data = data,
        estimation_type = "point",
        verbose = FALSE))
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )

  testthat::test_that(
    paste0(
      "6. familiarData for the ", outcome_type, " outcome can be created from ",
      "the one-feature, one-sample dataset and the model created using the one-feature data set."),
    {
      # Test with one-sample dataset
      fam_data <- familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_single_feature_one_sample_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE)
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )
  
  # Create a dataset using the bad dataset.
  data <- familiar:::test_create_bad_data(outcome_type = outcome_type)

  # Train a simple linear GLM using the bad dataset. The model will fail to
  # build.
  fam_model <- suppressWarnings(familiar:::test_train(
    data = data,
    cluster_method = "none",
    imputation_method = "simple",
    hyperparameter_list = list("sign_size" = familiar:::get_n_features(data)),
    learner = learner,
    time_max = 1832))

  testthat::test_that(
    paste0(
      "7. familiarData for the ", outcome_type, " outcome can be created from ",
      "good data and the NULL model created using the bad data set."), 
    {
      # Test with small, but good dataset.
      fam_data <- suppressWarnings(familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_small_good_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE))
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )

  testthat::test_that(
    paste0(
      "8. familiarData for the ", outcome_type, " outcome can be created from ",
      "single-sample data and the NULL model created using the bad data set."),
    {
      # Test with one-sample dataset.
      fam_data <- familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_one_sample_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE)
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )
  
  testthat::test_that(
    paste0(
      "9. familiarData for the ", outcome_type, " outcome can be created from ",
      "empty data and the NULL model created using the bad data set."), 
    {
      # Test with empty dataset
      fam_data <- familiar::as_familiar_data(
        object = fam_model,
        data = familiar:::test_create_empty_data(outcome_type = outcome_type),
        estimation_type = "point",
        verbose = FALSE)
      
      # Test that an familiarDataObject is created.
      testthat::expect_s4_class(fam_data, "familiarData")
    }
  )
}

familiar_data_creation_unit_test("count")
familiar_data_creation_unit_test("continuous")
familiar_data_creation_unit_test("binomial")
familiar_data_creation_unit_test("multinomial")
familiar_data_creation_unit_test("survival")