tests/testthat/test-group-sums.R

In capybara: Fast and Memory Efficient Fitting of Linear Models with High-Dimensional Fixed Effects

#' srr_stats (tests)
#' @srrstats {G5.2} Validates the correctness of group operations like summation and covariance using precomputed results.
#' @srrstats {RE3.3} Compares outputs of group-related calculations with reference implementations to ensure consistency.
#' @srrstats {RE4.3} Confirms robustness of grouped computations under different random seeds.
#' @srrstats {RE6.0} Ensures proper handling of input matrix dimensions and group structures in grouped summation functions.
#' @noRd
NULL

test_that("group_sums_* works", {
  set.seed(123)
  m <- matrix(rnorm(9), ncol = 3, nrow = 3)
  w <- matrix(rnorm(3), ncol = 1, nrow = 3)
  v <- matrix(rnorm(3), ncol = 1, nrow = 3)
  k <- 2L
  jlist <- list(1L, 2L)

  expect_equal(
    round(group_sums_(m, w, jlist), 3),
    # the matrix comes from: round(alpaca:::groupSums(m, w, jlist), 3)
    matrix(c(4.144, 4.872, -2.942), ncol = 1, nrow = 3)
  )

  expect_equal(
    round(group_sums_spectral_(m, w, v, k, jlist), 3),
    # the matrix comes from: round(alpaca:::groupSumsSpectral(m, w, v, k,
    # jlist), 3)
    matrix(c(0, 0, 0), ncol = 1, nrow = 3)
  )

  expect_equal(
    round(group_sums_var_(m, jlist)[, 1], 3),
    # the vector comes from: round(alpaca:::groupSumsVar(m, jlist)[, 1], 3)
    c(2.483, 2.644, -0.779)
  )

  expect_equal(
    round(group_sums_cov_(m, m, jlist)[, 1], 3),
    # the vector comes from: round(alpaca:::groupSumsCov(m, m, jlist)[, 1], 3)
    c(0, 0, 0)
  )
})