tests/testthat/test-struct_pi.R

In tsPI: Improved Prediction Intervals for ARIMA Processes and Structural Time Series

context("Testing struct_pi")

set.seed(123)
x <- rnorm(10)
test_that("bogus arguments throw error",{
  expect_error(struct_pi(x, type = "arima"))
  expect_error(struct_pi(x, xreg = 1))
  expect_error(struct_pi(x, n_ahead = -1))
  expect_error(struct_pi(x, level = 2))
  expect_error(struct_pi(x, median = "true"))
  expect_error(struct_pi(x, prior = "custom"))
  expect_error(struct_pi(x, prior = "custom", custom_prior = "f"))
  expect_error(struct_pi(x, nsim = 0))
})

test_that("output of struct_pi is of correct size and form",{
  y <- ts(x, start = 2000, frequency = 12)
  pred <- struct_pi(y, n_ahead = 5, nsim = 50)
  expect_identical(dim(pred), c(5L, 5L))
  expect_s3_class(pred, "mts")
  expect_s3_class(pred, "ts")
  expect_s3_class(pred, "matrix")
  expect_identical(frequency(pred), frequency(y))
  expect_identical(start(pred), end(y)+c(0, 1))
})

test_that("same seeds give same results",{
    set.seed(1)
    pred1 <- struct_pi(x, nsim = 50)
    set.seed(1)
    pred2 <- struct_pi(x, nsim = 50)
    expect_identical(pred1, pred2)
})

test_that("larger nsim gives smaller se",{
  set.seed(1)
  pred1 <- struct_pi(x, nsim = 50)
  set.seed(1)
  pred2 <- struct_pi(x, nsim = 100)
  expect_gt(as.numeric(pred1[, "se_upr"]), as.numeric(pred2[, "se_upr"]))
})

test_that("struct_pi gives same results each time",{
  set.seed(1)
  pred <- struct_pi(Nile, nsim = 50)
  expect_equal(pred[1,"median"], 800.20719, tol = 1e-4, check.attributes = FALSE)
  expect_equal(pred[1,"lwr"], 505.08147, tol = 1e-4, check.attributes = FALSE)
  expect_equal(pred[1,"upr"], 1092.89858, tol = 1e-4, check.attributes = FALSE)
  expect_equal(pred[1,"se_lwr"], 5.35309, tol = 1e-4, check.attributes = FALSE)
  expect_equal(pred[1,"se_upr"], 4.167766, tol = 1e-4, check.attributes = FALSE)

})