diff --git a/DESCRIPTION b/DESCRIPTION
index fc79e6496..68cfb6741 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -1,7 +1,7 @@
Type: Package
Package: datawizard
Title: Easy Data Wrangling and Statistical Transformations
-Version: 0.13.0.18
+Version: 0.13.0.19
Authors@R: c(
person("Indrajeet", "Patil", , "patilindrajeet.science@gmail.com", role = "aut",
comment = c(ORCID = "0000-0003-1995-6531")),
diff --git a/NEWS.md b/NEWS.md
index 6f85cf237..35e549ffa 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -39,6 +39,9 @@ CHANGES
* `data_summary()` also accepts the results of `bayestestR::ci()` as summary
function (#483).
+* `ranktransform()` has a new argument `zeros` to determine how zeros should be
+ handled when `sign = TRUE` (#573).
+
BUG FIXES
* `describe_distribution()` no longer errors if the sample was too sparse to compute
diff --git a/R/ranktransform.R b/R/ranktransform.R
index f48c07981..e463efde8 100644
--- a/R/ranktransform.R
+++ b/R/ranktransform.R
@@ -10,6 +10,9 @@
#' @param method Treatment of ties. Can be one of `"average"` (default),
#' `"first"`, `"last"`, `"random"`, `"max"` or `"min"`. See [rank()] for
#' details.
+#' @param zeros How to handle zeros. If `"na"` (default), they are marked as
+#' `NA`. If `"signrank"`, they are kept during the ranking and marked as zeros.
+#' This is only used when `sign = TRUE`.
#' @param ... Arguments passed to or from other methods.
#' @inheritParams extract_column_names
#' @inheritParams standardize.data.frame
@@ -19,8 +22,11 @@
#' @examples
#' ranktransform(c(0, 1, 5, -5, -2))
#'
-#' # Won't work
-#' # ranktransform(c(0, 1, 5, -5, -2), sign = TRUE)
+#' # By default, zeros are converted to NA
+#' suppressWarnings(
+#' ranktransform(c(0, 1, 5, -5, -2), sign = TRUE)
+#' )
+#' ranktransform(c(0, 1, 5, -5, -2), sign = TRUE, zeros = "signrank")
#'
#' head(ranktransform(trees))
#' @return A rank-transformed object.
@@ -38,6 +44,7 @@ ranktransform <- function(x, ...) {
ranktransform.numeric <- function(x,
sign = FALSE,
method = "average",
+ zeros = "na",
verbose = TRUE,
...) {
# no change if all values are `NA`s
@@ -45,6 +52,12 @@ ranktransform.numeric <- function(x,
return(x)
}
+ zeros <- match.arg(zeros, c("na", "signrank"))
+ method <- match.arg(
+ method,
+ c("average", "first", "last", "random", "max", "min")
+ )
+
# Warning if only one value and return early
if (insight::has_single_value(x)) {
if (is.null(names(x))) {
@@ -54,7 +67,13 @@ ranktransform.numeric <- function(x,
}
if (verbose) {
- insight::format_warning(paste0("Variable `", name, "` contains only one unique value and will not be normalized."))
+ insight::format_warning(
+ paste0(
+ "Variable `",
+ name,
+ "` contains only one unique value and will not be normalized."
+ )
+ )
}
return(x)
@@ -70,16 +89,31 @@ ranktransform.numeric <- function(x,
}
if (verbose) {
- insight::format_warning(paste0("Variable `", name, "` contains only two different values. Consider converting it to a factor."))
+ # nolint
+ insight::format_warning(
+ paste0(
+ "Variable `",
+ name,
+ "` contains only two different values. Consider converting it to a factor."
+ )
+ )
}
}
-
if (sign) {
- ZEROES <- x == 0
- if (any(ZEROES) && verbose) insight::format_warning("Zeros detected. These cannot be sign-rank transformed.")
- out <- rep(NA, length(x))
- out[!ZEROES] <- sign(x[!ZEROES]) * rank(abs(x[!ZEROES]), ties.method = method, na.last = "keep")
+ if (zeros == "na") {
+ out <- rep(NA, length(x))
+ ZEROES <- x == 0
+ if (any(ZEROES) && verbose) {
+ insight::format_warning("Zeros detected. These cannot be sign-rank transformed.") # nolint
+ }
+ out[!ZEROES] <- sign(x[!ZEROES]) * rank(abs(x[!ZEROES]),
+ ties.method = method,
+ na.last = "keep"
+ )
+ } else if (zeros == "signrank") {
+ out <- sign(x) * rank(abs(x), ties.method = method, na.last = "keep")
+ }
} else {
out <- rank(x, ties.method = method, na.last = "keep")
}
@@ -102,6 +136,7 @@ ranktransform.grouped_df <- function(x,
method = "average",
ignore_case = FALSE,
regex = FALSE,
+ zeros = "na",
verbose = TRUE,
...) {
info <- attributes(x)
@@ -143,6 +178,7 @@ ranktransform.data.frame <- function(x,
method = "average",
ignore_case = FALSE,
regex = FALSE,
+ zeros = "na",
verbose = TRUE,
...) {
# evaluate arguments
diff --git a/man/ranktransform.Rd b/man/ranktransform.Rd
index 44b0052c4..6bb7cfdc3 100644
--- a/man/ranktransform.Rd
+++ b/man/ranktransform.Rd
@@ -8,7 +8,14 @@
\usage{
ranktransform(x, ...)
-\method{ranktransform}{numeric}(x, sign = FALSE, method = "average", verbose = TRUE, ...)
+\method{ranktransform}{numeric}(
+ x,
+ sign = FALSE,
+ method = "average",
+ zeros = "na",
+ verbose = TRUE,
+ ...
+)
\method{ranktransform}{data.frame}(
x,
@@ -18,6 +25,7 @@ ranktransform(x, ...)
method = "average",
ignore_case = FALSE,
regex = FALSE,
+ zeros = "na",
verbose = TRUE,
...
)
@@ -33,6 +41,10 @@ ranktransform(x, ...)
\code{"first"}, \code{"last"}, \code{"random"}, \code{"max"} or \code{"min"}. See \code{\link[=rank]{rank()}} for
details.}
+\item{zeros}{How to handle zeros. If \code{"na"} (default), they are marked as
+\code{NA}. If \code{"signrank"}, they are kept during the ranking and marked as zeros.
+This is only used when \code{sign = TRUE}.}
+
\item{verbose}{Toggle warnings.}
\item{select}{Variables that will be included when performing the required
@@ -110,8 +122,11 @@ from the input data frame.
\examples{
ranktransform(c(0, 1, 5, -5, -2))
-# Won't work
-# ranktransform(c(0, 1, 5, -5, -2), sign = TRUE)
+# By default, zeros are converted to NA
+suppressWarnings(
+ ranktransform(c(0, 1, 5, -5, -2), sign = TRUE)
+)
+ranktransform(c(0, 1, 5, -5, -2), sign = TRUE, zeros = "signrank")
head(ranktransform(trees))
}
diff --git a/tests/testthat/test-ranktransform.R b/tests/testthat/test-ranktransform.R
index 5996ba4f3..c250215cc 100644
--- a/tests/testthat/test-ranktransform.R
+++ b/tests/testthat/test-ranktransform.R
@@ -46,6 +46,22 @@ test_that("signed rank works as expected", {
))))
})
+test_that("argument 'zeros' works", {
+ x <- c(-1, 0, 2, -3, 4)
+ expect_warning(
+ ranktransform(x, sign = TRUE),
+ "cannot be sign-rank"
+ )
+ expect_identical(
+ ranktransform(x, sign = TRUE, zeros = "signrank"),
+ c(-2, 0, 3, -4, 5)
+ )
+ expect_error(
+ ranktransform(x, sign = TRUE, zeros = "foo"),
+ "should be one of"
+ )
+})
+
test_that("ranktransform works with data frames", {
set.seed(123)
expect_snapshot(ranktransform(BOD))
@@ -58,9 +74,9 @@ test_that("ranktransform works with data frames (grouped data)", {
skip_if_not_installed("poorman")
set.seed(123)
- value1 <- sample(1:20, 9, replace = TRUE)
+ value1 <- sample.int(20, 9, replace = TRUE)
set.seed(456)
- value2 <- sample(1:20, 9, replace = TRUE)
+ value2 <- sample.int(20, 9, replace = TRUE)
test_df <- data.frame(
id = rep(c("A", "B", "C"), each = 3),
@@ -69,6 +85,7 @@ test_that("ranktransform works with data frames (grouped data)", {
stringsAsFactors = FALSE
)
+ # nolint start: nested_pipe_linter
expect_identical(
test_df %>%
poorman::group_by(id) %>%
@@ -81,6 +98,7 @@ test_that("ranktransform works with data frames (grouped data)", {
stringsAsFactors = FALSE
)
)
+ # nolint end
})
@@ -99,6 +117,7 @@ test_that("ranktransform works with data frames containing NAs (grouped data)",
stringsAsFactors = FALSE
)
+ # nolint start: nested_pipe_linter
expect_identical(
test_df %>%
poorman::group_by(id) %>%
@@ -111,6 +130,7 @@ test_that("ranktransform works with data frames containing NAs (grouped data)",
stringsAsFactors = FALSE
)
)
+ # nolint end
})
# select helpers ------------------------------