tests/testthat/test-cbea-internals.R

In qpmnguyen/CBEA: Competitive Balances for Taxonomic Enrichment Analysis in R

requireNamespace("tidyverse", quietly = TRUE)

# reference gmean and gmean rows ####
ref_gmean <- function(vec) {
  return(exp(mean(log(unclass(vec)))))
}

ref_gmeanrow <- function(matrix) {
  return(apply(matrix, 1, ref_gmean))
}

ref_ilr <- function(matrix, index) {
  p <- ncol(matrix)
  size <- length(index)
  scale <- sqrt(size * (p - size) / p)
  num <- ref_gmeanrow(matrix[, index])
  denom <- ref_gmeanrow(matrix[, -index])
  return(scale * log(num / denom))
}

# generate a fake data set
df_pos <- matrix(rpois(1000, 3), 100, 10) %>% replace(which(. == 0), 1)
df_neg <- replace(df_pos, sample(seq_len(length(df_pos)), size = 3), -2)
df_zero <- replace(df_pos, sample(seq_len(length(df_pos)), size = 3), 0)
index <- c(1:3)

###### Testing for get_raw_score function ###################################
# Testing correctness
test_that("Testing correctness for get_raw_score", {
  expect_equal(get_raw_score(df_pos, index), ref_ilr(df_pos, index))
})

test_that("get_raw_score should return error if there are zeroes or negatives", {
  expect_error(get_raw_score(df_zero, index))
  expect_error(get_raw_score(df_neg, index))
})

# Testing input output
test_that("Testing get_raw_score returns a vector", {
  expect_vector(get_raw_score(df_pos, index))
})

test_that("Testing if warning is displayed if data.frame is included", {
  expect_message(get_raw_score(as.data.frame(df_pos), index))
})

##### Testing for the estimate_distr function ####
# Testing for input output first
scores <- get_raw_score(df_pos, index)
test_that("Making sure estimate_distr returns a list", {
  expect_type(estimate_distr(scores, distr = "norm"), "list")
  expect_type(estimate_distr(scores, distr = "mnorm"), "list")
})

test_that("Making sure the names test_that are using is correct", {
  expect_named(estimate_distr(scores, distr = "norm"), c("mean", "sd", "loglik"))
  expect_named(estimate_distr(scores, distr = "mnorm"), c("mu", "sigma", "lambda", "loglik"))
})

test_that("Return error if have too much NAs", {
  expect_error(estimate_distr(rep(NA, 100)))
  expect_error(estimate_distr(c(rnorm(10), rep(NA, 90))))
})