Understanding the Kruskal Wallis Test
The Kruskal Wallis test is a nonparametric method for comparing three or more independent groups. It is useful when values are ordinal, skewed, or not safely modeled by normal distributions. Instead of comparing raw means, it compares ranked values from the combined data set.
When to Use It
Use this test when each group contains independent observations. The response should be measured on at least an ordinal scale. Group shapes should be reasonably similar when you want to compare medians. The method can still show general location differences, but interpretation becomes broader when shapes vary.
How the Calculator Helps
This calculator accepts several group lines and ranks all observations together. It handles ties using average ranks and applies a tie correction to the test statistic. It reports the H statistic, degrees of freedom, approximate p value, mean ranks, and effect sizes. These details help you move from raw samples to a structured statistical decision.
Post Hoc Comparison
A significant overall test only says that at least one group differs. It does not name the pair. The post hoc section uses Dunn style pairwise comparisons. Each pair receives a rank difference, standard error, z value, raw p value, adjusted p value, and decision. Adjustment methods help control false positives when many pairs are tested.
Practical Interpretation
Start with the overall p value. If it is greater than alpha, report no statistically significant overall difference. If it is less than or equal to alpha, review the adjusted pairwise results. Also compare mean ranks. A higher mean rank usually means larger observed values. Effect size gives context for practical importance.
Data Tips
Enter clean numeric values. Keep one group per line. Label lines with names before a colon when possible. Avoid mixing repeated measures with independent samples. Very small groups can give unstable post hoc results. Outliers are allowed, but they still influence ranks. Always connect the result to study design and subject knowledge.
Reporting Results
State the sample size for each group, the corrected H statistic, degrees of freedom, p value, and selected alpha. Then summarize significant adjusted pairs. Include the adjustment method, because raw pairwise p values can look stronger than protected results in final reports.