Skip to contents

Kruskal-Wallis rank sum test

Source: R/kruskal-test.R
KW_TEST.Rd

KW_TEST() tests whether the distribution of a continuous variable differs across the levels of one or more grouping variables. It is the rank-based, distribution-free analogue to one-way ANOVA. If KW_TEST is supplied within the lazy-loaded pipeline, supply KW_TEST as a function i.e. prepare_test(.test = KW_TEST).

Usage

KW_TEST(.var_id = NULL, .data = NULL, ...)

Arguments

.var_id

A variable mapper <var_id>. Currently supports x_by(). When supplied, the test executes immediately. If .var_id maps multiple grouping variables, one Kruskal-Wallis test runs per grouping variable against the same continuous variable.

.data

A data frame. Only used on the standalone path.

...

Additional arguments passed to the implementation. See the Supported variable mapper section for the full list per path.

Value

A cld_exec object (in statim::conclude()), a stat_infer_spec object, or a test_spec when .var_id = NULL. kwtest_def_xby's baseline returns a class_kw_test object by default; its pairwise variant instead returns a plain list (kw_test, comps) with its own print method. kwtest_def_on's baseline returns a plain list (statistic, df, p_value) with its own print method — neither path shares a class between them.

Details

H0: all samples come from the same distribution (equal medians, under the assumption of equal shape). H1: at least one sample is stochastically greater than another.

Supported variable mapper <var_id>s

  • x_by(): grouped Kruskal-Wallis test, with optional pairwise comparisons. See details from kwtest-xby.

  • on(): one-sample Kruskal-Wallis test via a compiled backend. See details from kwtest-on.

See also

kwtest-xby, class_kw_test, statim::via(), statim::conclude()

Examples

set.seed(123)
x = rcauchy(50, 1, 1.5)
g = sample(letters[1:5], size = 50, replace = TRUE)
KW_TEST(x_by(x, g))
#> -- Summary ---------------------------------------------------------------------
#> 
#> ────────────────────────────────
#>   vars  statistic  df  p_value  
#> ────────────────────────────────
#>    g      1.022    4    0.906   
#> ────────────────────────────────
#> 
#> 

# multiple grouping variables -> one test per grouping variable
# (confirm this call shape against your actual x_by() signature)
g2 = sample(c("control", "treatment"), size = 50, replace = TRUE)
KW_TEST(x_by(x, c(g, g2)))
#> -- Summary ---------------------------------------------------------------------
#> 
#> ────────────────────────────────
#>   vars  statistic  df  p_value  
#> ────────────────────────────────
#>    g      1.022    4    0.906   
#>    g2     5.986    1    0.014   
#> ────────────────────────────────
#> 
#>