tests/testthat/test-obj_hill.R

In GeoTox: Spatiotemporal Mixture Risk Assessment

test_that("spot check", {
  
  par <- c(2, 3, dt(0, df = 3, log = TRUE))
  log10_conc <- par[2]
  resp <- rep(par[1] / 2, 5)
  
  expect_equal(obj_hill(par, log10_conc, resp), 0)
  
  par <- c(2, 3, 4, dt(0, df = 4, log = TRUE))
  log10_conc <- par[2]
  resp <- rep(par[1] / 2, 7)
  
  expect_equal(obj_hill(par, log10_conc, resp), 0)
  
})

# Testing the obj-hill functions
test_that("2 and 3 parameter models give accurate likelihood", {

## hill function for simulation
  hill_fun <- function(x, alpha, beta, theta){
    alpha / (1 + 10^(beta * (theta - x)))
  }

## Creating regular parameters
  tp <- 1
  log10_ga <- 1
  gw <- 2
  err <- 1

## 2-parameter model

  #simulate dose-response
  params2 <- c(tp, log10_ga, err)
  log10_conc <- seq(0.1,10,length.out = 50)
  resp <- hill_fun(log10_conc, tp, log10_ga, gw)

  # 2 parameter hill model expectation
  val <- obj_hill(params2, log10_conc, resp)
  expect_equal(val, -100.201, tolerance = 0.001)

## 3-parameter model

  # 3 parameter hill model expectation
  params3 <- c(tp, log10_ga, gw, err)
  val <- obj_hill(params3, log10_conc, resp)
  expect_equal(val, -99.23463, tolerance = 0.001)


#### Edge Tests

# Really tiny parameter values

  #simulate dose-response
  params_tiny <- c(1e-10, 1e-10, 1e-10, 1e-10)

  expect_no_warning(obj_hill(params_tiny, log10_conc, resp))


  # Really large parameter values

  #simulate dose-response
  params_large <- c(1e10, 1e10, 1e10, 1e10)

  expect_no_warning(obj_hill(params_tiny, log10_conc, resp))
})