tests/testthat/test_deriv.R

In n-kall/priorsense: Prior Diagnostics and Sensitivity Analysis

library(priorsense)

test_that("powerscale_derivative gives 0 for uniform log_component", {
  expect_equal(
    priorsense::powerscale_derivative(
      x = seq(0, 1, 0.01),
      log_component = log(rep(
        1 / length(seq(0, 1, 0.01)),
        length(seq(0, 1, 0.01))
      )),
      quantity = "mean"),
    c(psens_mean = 0)
  )
  expect_equal(
    priorsense::powerscale_derivative(
      x = seq(0, 1, 0.01),
      log_component = log(rep(
        1 / length(seq(0, 1, 0.01)),
        length(seq(0, 1, 0.01))
      )),
      quantity = "sd"),
    c(psens_sd = 0)
  )
  expect_equal(
    priorsense::powerscale_derivative(
      x = seq(0, 1, 0.01),
      log_component = log(rep(
        1 / length(seq(0, 1, 0.01)),
        length(seq(0, 1, 0.01))
      )),
      quantity = "var"),
    c(psens_var = 0)
  )
})

test_that("powerscale_derivative gives warning if not using mean, sd or var", {
  expect_warning(
    priorsense::powerscale_derivative(
      x = seq(0, 1, 0.01),
      log_component = log(1 + seq(0, 1, 0.01)),
      quantity = "median"),
    "Power-scaling derivative for medians or quantiles is zero. Consider using powerscale_gradients instead."
  )
  expect_warning(
    priorsense::powerscale_derivative(
      x = seq(0, 1, 0.01),
      log_component = log(1 + seq(0, 1, 0.01)),
      quantity = "q95"),
    "Power-scaling derivative for medians or quantiles is zero. Consider using powerscale_gradients instead."
  )
  expect_warning(
    priorsense::powerscale_derivative(
      x = seq(0, 1, 0.01),
      log_component = log(1 + seq(0, 1, 0.01)),
      quantity = "mad"),
    "Power-scaling derivative for medians or quantiles is zero. Consider using powerscale_gradients instead."
  )
})