tests/testthat/test-covMat.R

In scPCA: Sparse Contrastive Principal Component Analysis

context("Test routines for generating covariance matrices")
library(tibble)
library(dplyr)
library(Matrix)
library(DelayedArray)

test_that(paste0("Routine can be applied to data.frame, tibble, matrix,",
                 " sparse matrices, and DelayedMatrix inputs"), {
  # make sure that all outputs agree
  cov_test <- cov(scale(background_df))
  expect_equal(covMat(background_df), cov_test)
  expect_equal(covMat(as_tibble(background_df)), cov_test)
  expect_equal(covMat(as.matrix(background_df)), cov_test)
  expect_equal(covMat(as(as.matrix(background_df), "dgCMatrix")), cov_test)
  expect_equivalent(as.matrix(covMat(DelayedArray(background_df))), cov_test)
})

test_that("Routine can center, scale, center and scale, or neither", {
  expect_silent(covMat(background_df, center = TRUE, scale = TRUE))
  expect_silent(covMat(background_df, center = TRUE, scale = FALSE))
  expect_silent(covMat(background_df, center = FALSE, scale = TRUE))
  expect_silent(covMat(background_df, center = FALSE, scale = FALSE))

  expect_silent(covMat(as_tibble(background_df), center = TRUE, scale = TRUE))
  expect_silent(covMat(as_tibble(background_df), center = TRUE, scale = FALSE))
  expect_silent(covMat(as_tibble(background_df), center = FALSE, scale = TRUE))
  expect_silent(covMat(as_tibble(background_df), center = FALSE, scale = FALSE))

  expect_silent(covMat(as.matrix(background_df), center = TRUE, scale = TRUE))
  expect_silent(covMat(as.matrix(background_df), center = TRUE, scale = FALSE))
  expect_silent(covMat(as.matrix(background_df), center = FALSE, scale = TRUE))
  expect_silent(covMat(as.matrix(background_df), center = FALSE, scale = FALSE))
  
  expect_silent(covMat(as(as.matrix(background_df), "dgCMatrix"),
                center = TRUE, scale = TRUE))
  expect_silent(covMat(as(as.matrix(background_df), "dgCMatrix"),
                       center = TRUE, scale = FALSE))
  expect_silent(covMat(as(as.matrix(background_df), "dgCMatrix"),
                       center = FALSE, scale = TRUE))
  expect_silent(covMat(as(as.matrix(background_df), "dgCMatrix"),
                       center = FALSE, scale = FALSE))
  
  expect_silent(covMat(DelayedArray(background_df),
                       center = TRUE, scale = TRUE))
  expect_silent(covMat(DelayedArray(background_df),
                       center = TRUE, scale = FALSE))
  expect_silent(covMat(DelayedArray(background_df),
                       center = FALSE, scale = TRUE))
  expect_silent(covMat(DelayedArray(background_df),
                       center = FALSE, scale = FALSE))
})

test_that(paste(
  "Variables with zero variance have zeros rows/columns in",
  "covariance matrix"
), {
  # add a constant column to the background data
  test_df <- background_df %>%
    mutate(
      zero_var = 1
    )
  expect_equal(sum(covMat(test_df)[, 31]), 0)
  expect_equal(sum(covMat(test_df)[31, ]), 0)
})

test_that("Always returns a matrix object", {
  expect_equal(class(covMat(background_df, center = TRUE, scale = TRUE)),
               c("matrix", "array"))
  expect_equal(class(covMat(as_tibble(background_df),
                            center = TRUE, scale = TRUE)),
               c("matrix", "array"))
  expect_equal(class(covMat(as.matrix(background_df),
                            center = TRUE, scale = TRUE)),
               c("matrix", "array"))
  expect_equal(class(covMat(as(as.matrix(background_df), "dgCMatrix"),
                       center = TRUE, scale = TRUE)),
               c("matrix", "array"))
  expect_equal(class(covMat(DelayedArray(background_df),
                       center = TRUE, scale = TRUE)),
               c("matrix", "array"))
})