tests/testthat/test_dist_ll.R

In poweRlaw: Analysis of Heavy Tailed Distributions

test_that("Testing log-likelihood function", {

  ##Discrete Power-law
  discrete_data = readRDS("discrete_data.RData")
  m = displ$new(discrete_data)
  m$setPars(2.58); m$setXmin(2)
  expect_equal(dist_ll(m), -9155.62809, tol = 1e-4)

  ##Discrete Log-normal
  x = c(1, 1)
  m = dislnorm$new(x)
  m$setPars(c(1, 1))

  l = (plnorm(1.5, 1, 1) - plnorm(0.5, 1, 1)) / (1 - plnorm(0.5, 1, 1))
  ll1 = 2 * log(l)
  expect_equal(dist_ll(m), ll1, tol = 1e-4)

  x = c(1, 1, 3, 4)
  m$setDat(x);  m$setXmin(2)
  ll3 = (plnorm(3.5, 1, 1) - plnorm(2.5, 1, 1)) / (1 - plnorm(1.5, 1, 1))
  ll4 = (plnorm(4.5, 1, 1) - plnorm(3.5, 1, 1)) / (1 - plnorm(1.5, 1, 1))
  ll = log(ll3) + log(ll4)
  expect_equal(dist_ll(m), ll, tol = 1e-4)

  ##Discrete Poisson
  x = c(1, 1, 3, 4)
  m = dispois$new(x); m$setPars(2)
  ll = log(prod(dpois(x, 2) / (1 - sum(dpois(0, 2)))))
  expect_equal(dist_ll(m), ll, tol = 1e-4)

  m$setXmin(2)
  ll = log(prod(dpois(3:4, 2)  /  (1 - sum(dpois(0:1, 2)))))
  expect_equal(dist_ll(m), ll, tol = 1e-4)

  ##Discrete Exponential
  x = c(1, 1)
  m = disexp$new(x)
  m$setPars(1)
  l = (pexp(1.5, 1, 1) - pexp(0.5, 1, 1)) / (1 - pexp(0.5, 1, 1))
  ll1 = 2 * log(l)
  expect_equal(dist_ll(m), ll1, tol = 1e-4)
  x = c(1, 1, 3, 4)
  m$setDat(x);  m$setXmin(2)
  ll3 = (pexp(3.5, 1, 1) - pexp(2.5, 1, 1)) / (1 - pexp(1.5, 1, 1))
  ll4 = (pexp(4.5, 1, 1) - pexp(3.5, 1, 1)) / (1 - pexp(1.5, 1, 1))
  ll = log(ll3) + log(ll4)
  expect_equal(dist_ll(m), ll, tol = 1e-4)

  #######################################
  #######################################
  #######################################
  ##CTN Power-law
  ctn_data = readRDS("ctn_data.RData")
  m = conpl$new(ctn_data)
  m$setPars(2.53282); m$setXmin(1.43628)
  expect_equal(dist_ll(m), -9276.4, tolerance = 0.00001);

  ##Lognormal
  x = c(1, 1)
  m = conlnorm$new(x); m$setPars(c(1, 1))

  ll1 = sum(log(dlnorm(x, 1, 1) / (1 - plnorm(1, 1, 1))))
  expect_equal(dist_ll(m), ll1, tol = 1e-4)

  x = c(1, 1, 3, 4)
  m$setDat(x)
  ll2 = sum(log(dlnorm(3:4, 1, 1) / (1 - plnorm(1, 1, 1))))
  expect_equal(dist_ll(m), ll1 + ll2, tol = 1e-4)

  m$setXmin(2)
  ll3 = sum(log(dlnorm(3:4, 1, 1) / (1 - plnorm(2, 1, 1))))
  expect_equal(dist_ll(m), ll3, tol = 1e-4)


  ##Exponential
  x = c(1, 1)
  m = conexp$new(x); m$setPars(1)

  ll1 = sum(log(dexp(x, 1) / (1 - pexp(1, 1))))
  expect_equal(dist_ll(m), ll1, tol = 1e-4)

  x = c(1, 1, 3, 4)
  m$setDat(x)
  ll2 = sum(log(dexp(3:4, 1) / (1 - pexp(1, 1))))
  expect_equal(dist_ll(m), ll1 + ll2, tol = 1e-4)

  m$setXmin(2)
  ll3 = sum(log(dexp(3:4, 1) / (1 - pexp(2, 1))))
  expect_equal(dist_ll(m), ll3, tol = 1e-4)



}
)