Skip to contents

Wilcoxon / Mann-Whitney Tests

Source: R/stat_infer.R
stat_wilcox_test.Rd

Nonparametric counterparts of the t-test: Wilcoxon signed-rank (one-sample and paired) and Mann-Whitney U (two-sample). Input shapes, .by slicing and output layout match stat_t_test().

Usage

stat_wilcox_test(
  data,
  .cols,
  group,
  .by = NULL,
  mu = 0,
  paired = FALSE,
  alternative = "two.sided",
  conf_level = 0.95,
  p_adjust = "none"
)

Arguments

data

A data frame.

.cols

<tidy-select> Numeric test variable(s). One column for one-/two-sample tests; exactly two columns when paired = TRUE.

group

<tidy-select> Optional grouping column with exactly two levels (the two groups to compare). Required for two-sample tests, forbidden for paired and one-sample tests.

.by

<tidy-select> Optional slice columns; the test is repeated separately within each slice.

mu

Null value of the mean (or mean difference).

paired

Logical; paired test. Requires exactly two .cols and no group (differences are taken row-wise).

alternative

"two.sided", "greater", or "less".

conf_level

Confidence level.

p_adjust

P-value adjustment (see stats::p.adjust) applied across the returned rows.

Value

A tibble with class stat_infer in the same layout as stat_t_test(), with statistic, p.value, effect (rank-biserial correlation) and no df column (NA). For one-sample tests the effect is the rank-biserial correlation of the signed ranks versus mu (positive proportion minus negative proportion); it is NA when all deviations from mu are zero.

Examples

stat_wilcox_test(sleep, .cols = extra, group = group)
#> Wilcoxon rank-sum / signed-rank test (two.sided) 
#> # A tibble: 1 × 16
#>   variable n_used n_dropped estimate estimate1 estimate2 statistic    df p.value
#> * <chr>     <int>     <int>    <dbl>     <dbl>     <dbl>     <dbl> <dbl>   <dbl>
#> 1 extra        20         0       NA        NA        NA      25.5    NA  0.0658
#> # ℹ 7 more variables: conf_low <dbl>, conf_high <dbl>, effect <dbl>,
#> #   method <chr>, alternative <chr>, p.adjusted <dbl>, sig <chr>
stat_wilcox_test(mtcars, .cols = mpg, mu = 20)
#> Wilcoxon rank-sum / signed-rank test (two.sided) 
#> # A tibble: 1 × 16
#>   variable n_used n_dropped estimate estimate1 estimate2 statistic    df p.value
#> * <chr>     <int>     <int>    <dbl>     <dbl>     <dbl>     <dbl> <dbl>   <dbl>
#> 1 mpg          32         0       NA        NA        NA       249    NA   0.786
#> # ℹ 7 more variables: conf_low <dbl>, conf_high <dbl>, effect <dbl>,
#> #   method <chr>, alternative <chr>, p.adjusted <dbl>, sig <chr>