tests/testthat/test-build_mhmm.R

In seqHMM: Mixture Hidden Markov Models for Social Sequence Data and Other Multivariate, Multichannel Categorical Time Series

# create test data
set.seed(123)
s <- 2:3
k <- length(s)
obs <- seqdef(matrix(sample(letters[1:5], 50, replace = TRUE), ncol = 10))

test_that("build_mhmm returns object of class 'mhmm'", {
  expect_error(
    model <- build_mhmm(obs, n_states = s),
    NA
  )
  expect_s3_class(
    model,
    "mhmm"
  )
  expect_error(
    build_mhmm(obs, initial_probs = simulate_initial_probs(s, k),
              emission_probs = simulate_emission_probs(s, 5, k),
              transition_probs = simulate_transition_probs(s, k)),
    NA
  )
})
test_that("build_mhmm errors with incorrect dims", {
  expect_error(
    build_mhmm(obs, initial_probs = simulate_initial_probs(s, k+1),
               emission_probs = simulate_emission_probs(s, 5, k),
               transition_probs = simulate_transition_probs(s, k)),
    "Unequal list lengths of 'transition_probs', 'emission_probs' and 'initial_probs'."
  )
  expect_error(
    build_mhmm(obs, initial_probs = simulate_initial_probs(rev(s), k),
               emission_probs = simulate_emission_probs(s, 5, k),
               transition_probs = simulate_transition_probs(s, k)),
    "Lengths of `initial_probs` does not match with the number of states."
  )
  expect_error(
    build_mhmm(obs, initial_probs = simulate_initial_probs(s, k),
               emission_probs = simulate_emission_probs(rev(s), 5, k),
               transition_probs = simulate_transition_probs(s, k)),
    "Number of rows in 'emission_probs' is not equal to the number of states."
  )
  expect_error(
    build_mhmm(obs, initial_probs = simulate_initial_probs(s, k),
               emission_probs = simulate_emission_probs(s, 4, k),
               transition_probs = simulate_transition_probs(s, k)),
    "Number of columns in 'emission_probs' is not equal to the number of symbols."
  )
  expect_error(
    build_mhmm(obs, initial_probs = simulate_initial_probs(s, k),
               emission_probs = simulate_emission_probs(s, 5, k)),
    "Provide either 'n_states' or all three of 'initial_probs', 'transition_probs', and 'emission_probs'."
  )
})

test_that("build_mhmm formula works", {
  expect_error(
    build_mhmm(obs, n_states = k, formula = ~ 1),
    "Argument 'n_states' is of length 1, leading to ordinary HMM. Please use 'build_hmm' instead."
  )
  expect_error(
    build_mhmm(obs, n_states = s, formula = ~ 1),
    "Argument 'data' is missing, but 'formula' was provided."
  )
  # default error message from model.matrix, could be more informative..
  expect_error(
    build_mhmm(obs, n_states = s, formula = ~ x, data = data.frame(y = 1)),
    "object 'x' not found"
  )
  expect_error(
    build_mhmm(obs, n_states = s, formula = ~ x, data = data.frame(x = 1)),
    paste0(
      "Number of subjects in data for covariates does not match the ",
      "number of subjects in the sequence data."
    )
  )
  expect_error(
    model <- build_mhmm(obs, n_states = s, formula = ~ x,
              data = data.frame(x = 1:5),
              cluster_names = c("A", "B")),
    NA
  )
  expect_equal(
    names(model$emission_probs),
    c("A", "B")
  )
  expect_equal(
    lengths(model$emission_probs),
    c(A = 10, B = 15)
  )
  expect_equal(
    model$symbol_names,
    c("a", "b", "c", "d", "e")
  )
})

test_that("build_mhmm returns the correct number of states", {
  expect_error(
    model <- build_mhmm(obs, n_states = s),
    NA
  )
  expect_equal(
    lengths(model$initial_probs),
    c("Cluster 1" = 2, "Cluster 2" = 3)
  )
  expect_equal(
    lapply(model$emission_probs, dim),
    list("Cluster 1" = c(2, 5), "Cluster 2" = c(3, 5))
  )
})