Nothing
# A `digits_*` argument may be a single number for the whole column, or one
# number per row. Before, it had to be a single number, so a data frame whose
# values were reported with different numbers of decimal places could not be
# tested correctly at all: any single `digits_x` was wrong for some of the rows.
# `0.1` at one decimal place is GRIM-consistent with n = 28, but the same value
# read as "0.10" is not. `4.71` needs two decimal places either way.
df_grim <- tibble::tibble(x = c(4.71, 0.1), n = c(28, 28))
df_grimmer <- tibble::tibble(
x = c(7.3, 5.23),
sd = c(2.51, 2.55),
n = c(12, 35)
)
df_debit <- tibble::tibble(
x = c(0.35, 0.53),
sd = c(0.18, 0.5),
n = c(20, 1683)
)
test_that("basic mappers take one `digits_*` value per row", {
# Testing each row on its own must give the same verdicts as testing them
# together with a per-row `digits_x`:
per_row <- df_grim |> grim_map(digits_x = c(2, 1), items_averaged = 1, data_type = "integer") |> purrr::pluck("consistency")
separate <- c(
df_grim[1, ] |> grim_map(digits_x = 2, items_averaged = 1, data_type = "integer") |> purrr::pluck("consistency"),
df_grim[2, ] |> grim_map(digits_x = 1, items_averaged = 1, data_type = "integer") |> purrr::pluck("consistency")
)
per_row |> expect_equal(separate)
# `0.1` at one decimal place is not the same test as `0.10` at two:
per_row |> expect_equal(c(TRUE, TRUE))
df_grim |>
grim_map(digits_x = 2, items_averaged = 1, data_type = "integer") |>
purrr::pluck("consistency") |>
expect_equal(c(TRUE, FALSE))
separate <- c(
df_grimmer[1, ] |> grimmer_map(digits_x = 1, digits_sd = 2, items_averaged = 1, data_type = "integer") |> purrr::pluck("consistency"),
df_grimmer[2, ] |> grimmer_map(digits_x = 2, digits_sd = 2, items_averaged = 1, data_type = "integer") |> purrr::pluck("consistency")
)
df_grimmer |>
grimmer_map(digits_x = c(1, 2), digits_sd = 2, items_averaged = 1, data_type = "integer") |>
purrr::pluck("consistency") |>
expect_equal(separate)
separate <- c(
df_debit[1, ] |> debit_map(digits_x = 2, digits_sd = 2, data_type = "integer") |> purrr::pluck("consistency"),
df_debit[2, ] |> debit_map(digits_x = 2, digits_sd = 1, data_type = "integer") |> purrr::pluck("consistency")
)
df_debit |>
debit_map(digits_x = 2, digits_sd = c(2, 1), data_type = "integer") |>
purrr::pluck("consistency") |>
expect_equal(separate)
})
test_that("a single `digits_*` value still applies to the whole column", {
per_row <- df_grim |> grim_map(digits_x = c(2, 2), items_averaged = 1, data_type = "integer") |> purrr::pluck("consistency")
df_grim |>
grim_map(digits_x = 2, items_averaged = 1, data_type = "integer") |>
purrr::pluck("consistency") |>
expect_equal(per_row)
})
test_that("a `digits_*` vector of the wrong length is an error", {
df_grim |>
grim_map(digits_x = c(2, 1, 2), items_averaged = 1, data_type = "integer") |>
expect_error("length 1 or the number of rows")
df_grimmer |>
grimmer_map(digits_x = c(1, 2, 1), digits_sd = 2, items_averaged = 1, data_type = "integer") |>
expect_error("length 1 or the number of rows")
})
# In a total-n mapper, a length-2 `digits_*` states the decimal places of the
# two groups. This is what makes the Bauer and Francis (2021) case study in
# `vignette("grim")` reproducible: it reports means of 4.71 and 5.3, which have
# different numbers of decimal places.
test_that("total-n mappers take one `digits_*` value per group", {
df <- tibble::tibble(x1 = 4.71, x2 = 5.3, n = 40)
out <- df |> grim_map_total_n(digits_x = c(2, 1), items_averaged = 1, data_type = "integer") |> audit()
# 17/23 with the original pairing, plus 19/21 and 16/24 with it reversed:
out$hits_forth |> expect_equal(1L)
out$hits_back |> expect_equal(2L)
out$hits_total |> expect_equal(3L)
# Testing 5.3 as "5.30" is stricter and finds only what Bauer and Francis did:
out_strict <- df |> grim_map_total_n(digits_x = 2, items_averaged = 1, data_type = "integer") |> audit()
out_strict$hits_total |> expect_equal(1L)
})
test_that("total-n mappers reject a `digits_*` vector that isn't length 2", {
df <- tibble::tibble(x1 = 4.71, x2 = 5.3, n = 40)
df |>
grim_map_total_n(digits_x = c(2, 1, 2), items_averaged = 1, data_type = "integer") |>
expect_error("must have length 1 or 2")
})
test_that("sequence mappers disperse each case on its own decimal level", {
out <- df_grim |> grim_map_seq(digits_x = c(2, 1), items_averaged = 1, data_type = "integer", dispersion = 1:2, include_consistent = TRUE)
out |> dplyr::filter(case == 1, var == "x") |> _$x |> expect_equal(c(4.69, 4.70, 4.72, 4.73))
out |> dplyr::filter(case == 2, var == "x") |> _$x |> expect_equal(c(-0.1, 0.0, 0.2, 0.3))
out |> dplyr::distinct(case, digits_x) |> _$digits_x |> expect_equal(c(2L, 1L))
out |> with(grim(x, n, digits_x, items_averaged, data_type)) |> expect_equal(out$consistency)
# The same from a column:
df_grim |> dplyr::mutate(digits_x = c(2, 1)) |> grim_map_seq(items_averaged = 1, data_type = "integer", dispersion = 1:2, include_consistent = TRUE) |> unclass_scr() |> expect_equal(unclass_scr(out))
# A case to disperse needs its decimal level:
df_grim |> dplyr::mutate(x = c(5.19, 0.1)) |> grim_map_seq(digits_x = c(NA, 1), items_averaged = 1, data_type = "integer") |> expect_error("missing in a case to disperse")
})
test_that("mappers take `digits_*` from columns of `data`", {
df_grim |> dplyr::mutate(digits_x = c(2, 1)) |> grim_map(items_averaged = 1, data_type = "integer") |> unclass_scr() |>
expect_equal(df_grim |> grim_map(digits_x = c(2, 1), items_averaged = 1, data_type = "integer") |> unclass_scr())
df_grimmer |> dplyr::mutate(digits_x = 1:2, digits_sd = 2) |> grimmer_map(items_averaged = 1, data_type = "integer") |> _$consistency |>
expect_equal(df_grimmer |> grimmer_map(digits_x = 1:2, digits_sd = 2, items_averaged = 1, data_type = "integer") |> _$consistency)
# In a total-n mapper, a row's value applies to both groups:
df <- tibble::tibble(x1 = c(4.71, 5.3), x2 = c(5.38, 4.9), n = c(40, 40))
out <- df |> dplyr::mutate(digits_x = c(2, 1)) |> grim_map_total_n(items_averaged = 1, data_type = "integer")
out |> dplyr::distinct(case, digits_x) |> _$digits_x |> expect_equal(c(2L, 1L))
out |> with(grim(x, n, digits_x, items_averaged, data_type)) |> expect_equal(out$consistency)
# Without either, the error says that a column would do:
df_grim |> grim_map(items_averaged = 1, data_type = "integer") |> expect_error("`digits_x` column")
})
# The dispersion step size has to come from `digits_*`, not from the stored
# value. A mean reported as 5.30 is stored as `5.3`, so going by the value gave
# it steps of 0.1 while a neighbouring 4.71 got steps of 0.01 -- in the same
# call, with the same `digits_x`.
test_that("`*_map_seq()` steps on the decimal level given by `digits_*`", {
out <- tibble::tibble(x = c(5.3, 4.71), n = c(40, 40)) |>
grim_map_seq(digits_x = 2, items_averaged = 1, data_type = "integer", dispersion = 1:3, include_consistent = TRUE)
out_x <- out[out$var == "x", ]
# Every dispersed value is one hundredth away from its neighbour:
out_x$x |> decimal_places() |> max() |> expect_equal(2L)
c(5.3, 4.71) |>
rep(each = 6) |>
call_on(\(v) out_x$x - v) |>
abs() |>
round(10) |>
call_on(\(x) x <= 0.03) |>
all() |>
expect_true()
# `n` has no `digits_n`, so it keeps stepping by whole numbers:
out$n[out$var == "n"] |>
call_on(\(x) x %% 1 == 0) |>
all() |>
expect_true()
})
test_that("an all-`NA` `digits_*` column of any type is an integer `NA`", {
df <- tibble::tibble(x = c(5.19, 4.2), n = c(28, 25))
for (na in list(NA, NA_character_, NA_real_)) {
out <- df |> dplyr::mutate(digits_x = na) |> grim_map(items_averaged = 1, data_type = "integer")
out$digits_x |> expect_identical(c(NA_integer_, NA_integer_))
out$consistency |> expect_identical(grim(df$x, df$n, NA, items_averaged = 1, data_type = "integer"))
}
out <- df |> grim_map(digits_x = NA, items_averaged = 1, data_type = "integer")
out$digits_x |> expect_identical(c(NA_integer_, NA_integer_))
})
test_that("a `digits_*` column or argument of strings is an error that names it", {
df <- tibble::tibble(x = c(5.19, 4.2), n = c(28, 25))
df |> dplyr::mutate(digits_x = "2") |> grim_map(items_averaged = 1, data_type = "integer") |> expect_error("`digits_x` column of `data` is a character vector")
df |> grim_map(digits_x = "2", items_averaged = 1, data_type = "integer") |> expect_error("`digits_x` must be numeric")
df |> dplyr::mutate(digits_x = factor("2")) |> grim_map(items_averaged = 1, data_type = "integer") |> expect_error("not a factor")
# One bad value among numbers is still blamed on its row:
df |> grim_map(digits_x = c(2, 1.5), items_averaged = 1, data_type = "integer") |> expect_error("In row 2")
})
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.