📊 Kruskal-Wallis H Test Calculator
Paste your groups (one per line) to run a Kruskal-Wallis test — the rank-based alternative to ANOVA that doesn't assume normal data.
H statistic
9.981
Degrees of freedom
2
p-value
0.0068
Rank sums: 34 · 64.5 · 21.5 · N = 15, 3 groups
At least one group differs (significant at α = 0.05)
p = 0.0068 < 0.05, so you reject the null hypothesis that all 3 groups have the same distribution. Like ANOVA, it doesn't say which groups differ — use a post-hoc test (e.g. Dunn's test).
The Kruskal-Wallis H test is the non-parametric alternative to one-way ANOVA: it compares three or more groups by ranking all values together, so it doesn't assume the data are normally distributed — useful for skewed data, ordinal ratings or outliers. The H statistic is compared to a chi-square distribution (with a tie correction). A significant result means the groups aren't all alike but not which ones differ. 🔒 In your browser.
How the kruskal-wallis h test calculator works
The Kruskal-Wallis test ranks every value across all groups together, sums the ranks within each group, and forms the H statistic from those rank sums. H is compared to a chi-square distribution with k − 1 degrees of freedom (k = number of groups), and the tool applies a tie correction for repeated values. Enter one group per line, values separated by commas or spaces.
This is the test to reach for when you'd use one-way ANOVA but the data isn't normally distributed — skewed measurements, ordinal ratings or small samples with outliers — because it works on ranks. Like ANOVA, a significant H tells you the groups are not all alike but not which ones differ; a post-hoc procedure such as Dunn's test locates the differences. For exactly two groups it reduces to the Mann-Whitney U test. It all runs locally in your browser.
Frequently asked questions
When should I use the Kruskal-Wallis test?
When comparing three or more independent groups but the normality assumption of ANOVA is doubtful — skewed data, ordinal scales, small samples or outliers. It's the non-parametric alternative to one-way ANOVA and works on ranks.
How does it relate to ANOVA and Mann-Whitney?
It generalises the Mann-Whitney U test (two groups) to three or more, and it plays the role ANOVA plays for normal data. Use ANOVA when the groups are roughly normal, Kruskal-Wallis when they're not.
What does a significant H mean?
That the groups don't all come from the same distribution — at least one differs. It doesn't identify which; follow up with a post-hoc test like Dunn's test with a multiple-comparison correction.
Does it handle tied values?
Yes — tied values receive average ranks and the H statistic is adjusted with a tie correction, so repeated numbers are handled correctly.
Is my data uploaded?
No — the test is computed entirely in your browser and works offline, so your data never leaves your device.