tests/testthat/test-plot-retention.R

In EFAtools: Fast and Flexible Implementations of Exploratory Factor Analysis Tools

test_that("efa_retention plot methods return ggplot objects", {
  skip_if_not_slow()
  ekc <- EKC(test_models$baseline$cormat, N = 500)
  p_ekc <- plot(ekc)
  expect_s3_class(p_ekc, "ggplot")
  expect_equal(ekc$results[[1]]$plot_type, "eigen")

  hull <- HULL(test_models$baseline$cormat, N = 500, method = "ML")
  p_hull <- plot(hull)
  expect_s3_class(p_hull, "ggplot")
  expect_equal(hull$results[[1]]$plot_type, "hull")

  kgc <- KGC(test_models$baseline$cormat)
  p_kgc <- plot(kgc)
  expect_s3_class(p_kgc, "ggplot")
  expect_equal(.retention_record(kgc, "PCA")$plot_type, "eigen")

  # CD is stochastic, so only smoke-test that the plot builds with its custom
  # y-axis label (no vdiffr baseline)
  set.seed(123)
  cd <- CD(GRiPS_raw, N_pop = 1000, N_samples = 100)
  p_cd <- plot(cd)
  expect_s3_class(p_cd, "ggplot")
  expect_equal(.retention_record(cd, "CD")$y_label, "RMSE eigenvalues")

  scree <- SCREE(test_models$baseline$cormat)
  p_scree <- plot(scree)
  expect_s3_class(p_scree, "ggplot")

  # PARALLEL: smoke-test only (no vdiffr baseline) because the simulated
  # reference lines vary with the RNG state and the future plan's chunking,
  # so an SVG baseline would not be portable
  pa <- PARALLEL(test_models$baseline$cormat, N = 500, eigen_type = c("PCA", "SMC"))
  p_pa <- plot(pa)
  expect_s3_class(p_pa, "ggplot")

  # no real data: the plot shows only the dashed reference series
  pa_nodat <- PARALLEL(N = 20, n_vars = 5, eigen_type = "PCA")
  expect_s3_class(plot(pa_nodat), "ggplot")
})

test_that("plot-less criteria return NULL with a message", {
  for (obj in list(MAP(test_models$baseline$cormat),
                   SMT(test_models$baseline$cormat, N = 500))) {
    expect_message(p <- plot(obj), class = "efa_no_plot")
    expect_null(p)
  }
})

test_that("NEST plots its empirical eigenvalues against the reference series", {
  # few simulated datasets: the plot only needs a well-formed record
  set.seed(42)
  nest <- efa_nest(test_models$baseline$cormat, N = 500, n_datasets = 50)

  expect_s3_class(plot(nest), "ggplot")

  rec <- .retention_record(nest, "NEST")
  expect_equal(rec$plot_type, "eigen")
  # the shared eigenvalue plotter binds the series into one data frame, so the
  # reference has to be as long as the empirical eigenvalues
  expect_length(rec$reference, length(rec$y))
  # the retained solution is marked, as in the other eigenvalue plots (the record
  # carries no highlight when nothing is retained, so pin that there is one)
  expect_gte(nest$n_factors[["NEST"]], 1)
  expect_equal(rec$highlight, nest$n_factors[["NEST"]])
})

test_that("eigen plot of an empty record returns NULL with a message", {
  # e.g. CD on a tiny dataset that suggests 0 factors -> empty x/y record
  obj <- .new_efa_retention(
    "CD",
    results = list(list(name = "CD", label = "Suggested number of factors",
                        n_factors = 0, plot_type = "eigen",
                        x = integer(0), y = numeric(0))),
    settings = list()
  )
  expect_message(p <- plot(obj), class = "efa_no_plot")
  expect_null(p)
})

test_that("efa_retain reports when none of its criteria has a plot", {
  # plot.efa_retain() drops the per-criterion NULLs; with MAP and SMT the only
  # criteria run, nothing is left to return.
  nf <- efa_retain(test_models$baseline$cormat, N = 500, suitability = FALSE,
                   criteria = c("MAP", "SMT"))
  expect_message(p <- plot(nf), class = "efa_no_plot")
  expect_null(p)
})

test_that("EKC eigen plot is visually stable", {
  skip_if_not_installed("vdiffr")

  ekc <- EKC(test_models$baseline$cormat, N = 500)
  vdiffr::expect_doppelganger("EKC eigen plot", plot(ekc))
})

test_that("HULL hull plot is visually stable", {
  skip_if_not_installed("vdiffr")

  hull <- HULL(test_models$baseline$cormat, N = 500, method = "ML")
  vdiffr::expect_doppelganger("HULL hull plot", plot(hull))
})

test_that("KGC eigen plot is visually stable", {
  skip_if_not_installed("vdiffr")

  kgc <- KGC(test_models$baseline$cormat)
  vdiffr::expect_doppelganger("KGC eigen plot", plot(kgc))
})

test_that("SCREE eigen plot is visually stable", {
  skip_if_not_installed("vdiffr")

  scree <- SCREE(test_models$baseline$cormat)
  vdiffr::expect_doppelganger("SCREE eigen plot", plot(scree))
})

test_that("plot.efa_retain's deterministic members are visually stable", {
  skip_if_not_installed("vdiffr")

  # plot.efa_retain() dispatches one plot per plottable criterion that was run. Only the
  # deterministic ones carry a baseline: CD, PARALLEL, and NEST simulate, so their
  # reference series depend on the RNG state and no SVG baseline would be portable.
  nf <- efa_retain(test_models$baseline$cormat, N = 500, suitability = FALSE,
                   criteria = c("EKC", "KGC", "SCREE"), eigen_type_other = "PCA")
  p <- plot(nf)

  expect_named(p, c("EKC", "KGC", "SCREE"))
  for (id in names(p)) {
    vdiffr::expect_doppelganger(paste("efa_retain", id, "plot"), p[[id]])
  }
})