tests/testthat/test_univariateML.R

In kdensity: Kernel Density Estimation with Parametric Starts and Asymmetric Kernels

univariateML_densities = intersect(
  names(starts),
  univariateML::univariateML_models
  )

# There are some exceptions we won't test.

exceptions = c("pareto", "unif", "lomax")

univariateML_densities = setdiff(
  univariateML_densities,
  exceptions
  )

generate_random = function(n, support) {
  if (support == "c(-Inf, Inf)") return(rnorm(n))
  if (support == "c(0, Inf)") return(rgamma(n, 4, 4))
  if (support == "c(1, Inf)") return(rexp(n) + 1)
  if (support == "c(0, 1)") return(runif(n))
  stop()
}

# Actual testing for all non-exceptions.
set.seed(10)
n = 50

for(name in univariateML_densities) {
  fun = eval(parser(paste0("univariateML::ml", name)))
  support = deparse(get_density_and_support(fun)$support)
  rands = generate_random(n, support)
  kde = kdensity(rands, start = name)
  coef(kde)
  logLik(kde)
  AIC(kde)
}

# For Lomax
rands = extraDistr::rlomax(n, 2, 3)
kde = kdensity(rands, start = "lomax")
coef(kde)
logLik(kde)
AIC(kde)