Pairwise post-hoc test following Kruskal-Wallis test

This is a tidy wrapper around dunn.test::dunn.test(). It performs Dunn's test, a non-parametric pairwise follow-up for a Kruskal-Wallis test.

Usage

dunn_test(
  x,
  groups,
  data,
  p.adjust.method = c("holm", "hochberg", "bonferroni", "bh", "by", "sidak", "hs",
    "none"),
  alpha = 0.05
)

Arguments

x

a numeric vector.

groups

a vector or factor giving the group for the corresponding elements of x.

data

a data frame containing the variables.

p.adjust.method

character. method for adjusting p-values for multiple comparisons.

alpha

numeric. alpha level.

Value

A tibble with columns:

• contrast: The compared groups.

• statistic: Dunn's test statistic (z).

• adj.p.value: Adjusted p-value based on the specified p.adjust.method.

See also

dunn.test::dunn.test()

Examples

mtcars2 <- transform(mtcars, cyl = factor(cyl))
kruskal.test(mpg ~ gear, data = mtcars2)
#> 
#> 	Kruskal-Wallis rank sum test
#> 
#> data:  mpg by gear
#> Kruskal-Wallis chi-squared = 14.323, df = 2, p-value = 0.0007758
#> 
dunn_test(mpg, gear, data = mtcars2)
#> # A tibble: 3 × 3
#>   contrast statistic adj.p.value
#>   <chr>        <dbl>       <dbl>
#> 1 3 - 4        -3.76    0.000253
#> 2 3 - 5        -1.65    0.0998  
#> 3 4 - 5         1.14    0.127