tests/testthat/test_Roc.R

In easyVerification: Ensemble Forecast Verification for Large Data Sets

library(easyVerification)
context("ROC scores")

obs <- diag(1, 4)[-4, ]
fcst <- obs
fcst2 <- obs * 0.2 + 0.2
fcst3 <- array(0.25, dim(obs))
obs2 <- rbind(diag(1, 4)[-4, ], diag(1, 4)[-3, ], diag(1, 4)[-1, ])
fcst4 <- pmax(obs2 + runif(length(obs2)), 0)
fcst4 <- fcst4 / rowSums(fcst4)

test_that("ROC area values", {
  for (use.easy in c(FALSE, TRUE)) {
    expect_is(EnsRoca(fcst, obs, use.easy = use.easy), "list")
    expect_equal(length(EnsRoca(fcst, obs, use.easy = use.easy)), 8)
    expect_equal(EnsRoca(fcst, obs, use.easy = use.easy)[1:4], list(cat1 = 1, cat2 = 1, cat3 = 1, cat4 = as.numeric(NA)))
    expect_equal(EnsRoca(fcst3, obs, use.easy = use.easy)[1:4], list(cat1 = 0.5, cat2 = 0.5, cat3 = 0.5, cat4 = as.numeric(NA)))
    expect_equal(EnsRoca(fcst, obs, use.easy = use.easy), EnsRoca(fcst2, obs, use.easy = use.easy))
  }
  expect_equal(EnsRoca(fcst4, obs2)[1:4], EnsRoca(fcst4, obs2, use.easy = TRUE)[1:4])
})