Analyses

Kruskal–Wallis H test

Compare three or more independent groups by their ranks.

Edit on GitHub

When to use it

Use Kruskal–Wallis instead of a one-way ANOVA when you have three or more groups and the scores are ordinal or skewed.

Assumptions

Groups are independent. The outcome is at least ordinal. The test does not tell you which groups differ until you turn on post hoc.

Running it in Tensr

Analyze → Nonparametric Tests → K Independent Samples. In chat: “Kruskal–Wallis of score by method.”
Group column and value column are required. Post hoc starts off.

Options

Prop

Type

Reading the output

The three teaching methods, compared on ranks. The same moderate mean gap as the ANOVA, read without assuming a normal curve.

GroupnMedianStatus
Lecture3269.1included
Online3271.05included
Workshop3272.95included

Kruskal-Wallis test of score by method (3 groups) on 96 rows. Effect size ε² = 0.076 (small; ε² < 0.01 negligible, 0.01 ≤ ε² < 0.08 small, 0.08 ≤ ε² < 0.26 medium, ε² ≥ 0.26 large). Cutoffs: Mangiafico, 2016. The primary result is significant (p = .027). This is large enough, in this sample, to treat the comparison this page is about as a real association rather than noise. Metrics: H statistic = 7.22; p-value = .027; ε² = 0.076; Effect size = small.

Reporting (APA 7)

Kruskal-Wallis test of score by method (3 groups) on 96 rows. Effect size ε² = 0.076 (small; ε² < 0.01 negligible, 0.01 ≤ ε² < 0.08 small, 0.08 ≤ ε² < 0.26 medium, ε² ≥ 0.26 large). Cutoffs: Mangiafico, 2016. This result is significant (p = .027). The effect is moderate, so report its size with the p-value.

Coming from SPSS

Analyze → Nonparametric Tests → K Independent Samples.