tests/testthat/test_batmix_fitting.R

In keesmulder/flexcircmix: Fit Mixtures of Flexible Circular Distributions

library(flexcircmix)

context("General tests for fitting batmix through wrapper fitbatmix")


test_that("EM fitbatmix wrapper works", {

  set.seed(15)

  x <- rinvbatmix(100)

  (system.time(fit_inv <- fitbatmix(x, bat_type = "inverse",
                                    optimization_its = 2,
                                    max_its = 2, method = "EM")))

  (system.time(fit_pow <- fitbatmix(x, bat_type = "power",
                                    optimization_its = 2,
                                    max_its = 2, method = "EM")))



  expect_true(fit_inv$method == "EM")
  expect_true(fit_pow$method == "EM")




  # Works with a single component
  expect_error(
    fit_pow <- fitbatmix(x, n_comp = 1, verbose = FALSE,
                         bat_type = "power", method = "EM")
    , NA)


  expect_equal(ncol(fit_pow$estimates), 6)


  x_mis <- x
  x_mis[c(20, 24)] <- NA

  expect_warning(fitbatmix(x_mis, bat_type = "power",
                                      optimization_its = 2,
                                      max_its = 2, method = "EM"))
})

test_that("Plotting works" , {

  x <- rinvbatmix(100)

  fit_pow <- fitbatmix(x, n_comp = 3, verbose = FALSE,
                       bat_type = "power", method = "EM")
  fit_mcmc <- fitbatmix(x, n_comp = 3, verbose = 0,
                        bat_type = "power", Q = 10, method = "bayes")

  plot_batmixfit(x, fit_pow$estimates)

  plot_batmix_sample(x, fit_mcmc$mcmc_sample, plot_n = 2)


  fit_pow <- fitbatmix(x, n_comp = 1, verbose = FALSE,
                       bat_type = "power", method = "EM")
  fit_mcmc <- fitbatmix(x, n_comp = 1, verbose = 0,
                        bat_type = "power", Q = 10, method = "bayes")

  expect_true(class(plot_batmixfit(x, fit_pow$estimates))[2] == "ggplot")
  expect_true(class(plot_batmix_sample(x, fit_mcmc$mcmc_sample, plot_n = 2))[2] == "ggplot")

})

test_that("Fixing parameters works", {

  x <- rinvbatmix(50)

  fixed_pmat = matrix(c(NA, NA, .3, NA,
                        1, NA, NA, NA,
                        NA, 2, NA, NA),
                      nrow = 3, ncol = 4, byrow = TRUE)

  fit_pow <- fitbatmix(x, n_comp = 3, verbose = FALSE,
                       fixed_pmat = fixed_pmat,
                       bat_type = "power", method = "EM")


  fit_mcmc <- fitbatmix(x, n_comp = 3, verbose = 0, bat_type = "power",
                       fixed_pmat = fixed_pmat,
                        Q = 10, method = "bayes")

  expect_true(fit_pow$estimates[1, 3] == fixed_pmat[1, 3])
  expect_true(fit_pow$estimates[2, 1] == fixed_pmat[2, 1])
  expect_true(fit_pow$estimates[3, 2] == fixed_pmat[3, 2])
  expect_true(fit_pow$estimates[1, 3] == fixed_pmat[1, 3])
  expect_true(fit_pow$estimates[2, 1] == fixed_pmat[2, 1])
  expect_true(fit_pow$estimates[3, 2] == fixed_pmat[3, 2])
})


test_that("Bootstrap works", {

  x <- rinvbatmix(200)

  # Test without parallelization
  expect_error(fit_boot <- fitbatmix(x, method = "boot", B = 3,
                                     parallel = FALSE), NA)
  expect_error(fit_boot, NA)
  expect_error(summary(fit_boot), NA)

  expect_equal(ncol(fit_boot$estimates), 6)





})

test_that("MCMC works", {

  x <- rinvbatmix(200)
  fit_mcmc <- fitbatmix(x, method = "bayes", Q = 10)
  fit_mcmc$mcmc_summary

  expect_equal(ncol(fit_mcmc$estimates), 6)
})