tests/testthat/test-model_eval.R

In LSTbook: Data and Software for "Lessons in Statistical Thinking"

test_that("model_eval() repeats input values and adds columns named .output, .lwr, .upr", {
  mod <- mtcars |> model_train(mpg ~ wt + hp)
  Tmp <- model_eval(mod, wt = c(2, 4), hp=c(100,200,300))
  expect_equal(c("wt", "hp", ".lwr", ".output", ".upr"), names(Tmp))
})

test_that("model_eval() handles ... inputs to assemble an input data set", {
  mod <- mtcars |> model_train(mpg ~ wt + hp)
  Tmp <- model_eval(mod, wt = c(2, 4), hp=c(100,200,300))
  expect_equal(nrow(Tmp), 6)
})

test_that("model_eval() uses training data as input if no other is specified", {
  mod <- mtcars |> model_train(mpg ~ wt + hp)
  Tmp <- suppressMessages(model_eval(mod))
  expect_equal(nrow(Tmp), nrow(mtcars))
})

test_that("model_eval() generates a skeleton of inputs when requested", {
  mod <- mtcars |> model_train(mpg ~ wt + hp)
  Tmp <- model_eval(mod, skeleton = TRUE, ncont=5)
  expect_true(length(unique(Tmp$wt)) >= 5)
  expect_true(length(unique(Tmp$hp)) >= 5)
  expect_true(nrow(Tmp) > 25)
})

test_that("model_eval() deals with transforms when generating a skeleton of inputs", {
  mod <- mtcars |> model_train(mpg ~ poly(wt,2) + splines::ns(hp,2))
  Tmp <- model_eval(mod, skeleton = TRUE, ncont=5)
  expect_true(length(unique(Tmp$wt)) >= 5)
  expect_true(length(unique(Tmp$hp)) >= 5)
  expect_true(nrow(Tmp) > 25)
})

test_that("model_eval() only outputs response and explanatory variables in the model, no matter what else is in the training data frame.", {
  mod <- mtcars |> model_train(hp ~ mpg)
  Res <- model_eval(mod) |> select(-".output", -".lwr", -".upr", -".resid")
  expect_true(all(names(Res) %in% c("hp", "mpg")))
})