tests/testthat/test-gaussian.R

In jolars/SLOPE: Sorted L1 Penalized Estimation

test_that("unregularized gaussian models work as expected", {
  set.seed(1)

  x <- as.matrix(abalone$x)
  y <- abalone$y

  lm_fit <- lm(y ~ as.matrix(x))

  g <- SLOPE(x, y, family = "gaussian", alpha = 0)

  expect_equivalent(
    coef(lm_fit),
    as.matrix(coef(g)),
    tol = 1e-3
  )
})

test_that("wide and tall inputs work correctly", {
  set.seed(926)

  grid <- expand.grid(
    n = c(50, 100),
    p = c(50, 100),
    density = c(1, 0.5)
  )

  for (i in seq_len(nrow(grid))) {
    n <- grid$n[i]
    p <- grid$p[i]
    d <- grid$density[i]

    xy <- SLOPE:::randomProblem(n, p, density = d)

    expect_silent(SLOPE(xy$x, xy$y, path_length = 5))
  }
})

test_that("diagonal X, known solution", {
  n <- p <- 4

  x <- diag(n)
  y <- c(8, 6, 4, 2)
  lambda <- c(4, 3, 2, 1)

  res <- SLOPE(
    x,
    y,
    family = "gaussian",
    intercept = FALSE,
    center = FALSE,
    scale = "none",
    lambda = lambda / n,
    alpha = 1
  )

  beta <- coef(res)

  expect_equal(as.vector(beta), c(4, 3, 2, 1), check.attributes = FALSE) })