tests/testthat/test-helper_estimates.R

In dustinfife/flexplot: Graphically Based Data Analysis Using 'flexplot'

context("helper_estimates")

test_that("return_levels_or_mean", {
  expect_equal(return_levels_or_mean("a", c("a"), small)%>%as.character, "a")
  expect_equal(return_levels_or_mean("z", "a", small), .3853, tol=.01)
})

test_that("return_averages works", {
  model = lm(y~a + z +b , data=small)
  expect_equal(return_averages(model, "a", TRUE)$a%>%as.character, "a")
  expect_equal(return_averages(model, "z", TRUE)$z, .3853, tol=.01)
  expect_message(return_averages(model, "a", FALSE))
  expect_null(return_averages(model, character(0)))
})

test_that("return_factor_levels works", {
  expect_equal(return_factor_levels("a", small)$a%>%as.character, c("a", "b"))
  expect_null(return_factor_levels(character(0), small))  
  expect_equal(return_factor_levels(c("a", "b"), small)$b%>%as.character, c("z", "y", "x"))
})

test_that("return_plus_minus_one_sd works", {
  expect_equal(return_plus_minus_one_sd(1:10)[1], 8.52765, tol=.01)
  expect_equal(length(return_plus_minus_one_sd(1:10)), 2)
})

test_that("anchor.predictions works for categorical predictors", {
  linear.model = lm(weight.loss~health + gender, data= exercise_data)
  expect_message(anchor.predictions(linear.model, "gender")$prediction[1])
  linear.model = lm(weight.loss~therapy.type + gender, data= exercise_data)
  expect_equal(anchor.predictions(linear.model, c("therapy.type", "gender"))$prediction[1],
               4.3475, tolerance = .002)
})

test_that("bf.bif works", {
  expect_equal(bf_bic(lm(y~x + a, data=small), lm(y~x * a, data=small)), 3.339977, tol = .01)
  expect_equal(bf_bic(lm(y~x + a, data=small), lm(y~x * a, data=small), invert=T), 1/3.339977, tol = .01)
})

test_that("which_terms_are_factors_or_numbers works", {
  expect_true(length(which_terms_are_factors_or_numbers(small, "a")$factors)==1)
  expect_true(length(which_terms_are_factors_or_numbers(small, "x")$factors) ==0)
  expect_true(length(which_terms_are_factors_or_numbers(small, c("x", "y", "a", "b"))$factors) == 2)
})

test_that("compute_semi_partial works", {
  mod = lm(y~x, data=small)
  expect_equal(as.numeric(compute_semi_partial(mod)), summary(mod)$r.squared)
  expect_equal(names(compute_semi_partial(lm(y~a, data=small))), "a")
  expect_equal(names(compute_semi_partial(lm(y~a+b+z, data=small))), c("a", "b", "z"))
  expect_message(compute_semi_partial(lm(y~a*b, data=small)))
})

test_that("populate_estimates_factors works", {
  expect_equal(populate_estimates_factors(lm(y~x, data=small))$coef.matrix, NA)
  expect_equal(levels(populate_estimates_factors(lm(y~a, data=small))$coef.matrix$variables)[2], "a")
})

test_that("populate_estimates_numeric works", {
  expect_equal(dim(populate_estimates_numeric(lm(y~x, data=small))), c(2,7))
  expect_equal(dim(populate_estimates_numeric(lm(y~a, data=small))), NULL)
})

test_that("compute_r_squared works", {
  expect_equal(length(compute_r_squared(lm(y~x, data=small))), 3)
  expect_true(all(compute_r_squared(lm(y~1, data=small))==0))
})

test_that("compute_correlation works", {
  expect_true(is.numeric(compute_correlation(lm(y~x, data=small))))
  expect_true(is.na(compute_correlation(lm(y~a, data=small))))
  expect_true(is.na(compute_correlation(lm(y~x + z, data=small))))
})

test_that("bic works", {
  model.me = lm(weight.loss ~ motivation+therapy.type, data = exercise_data)
  model.int = lm(weight.loss ~ motivation*therapy.type, data = exercise_data)
  expect_equal(bf.bic(model.me, model.int, invert=T), .01049, tolerance=.001)
})  

test_that("icc works", {
  data(math)
  mod = lmer(MathAch~1 + (1|School), data=math)
  expect_equal(icc(mod)$icc, .18, tol=.01)
})

test_that("removing interactions works", {
  mod = lm(kills~agility*speed, data=avengers)
  expect_true(length(remove_interaction_terms(mod))==2)
})

test_that("generate_grid_predictions", {
  expect_equal(nrow(generate_grid_predictions(list(a=1:3, b=4:5), list(c = c('a', 'b')), NULL)), 12)
  expect_null(generate_grid_predictions(NULL, NULL, NULL))
})