Kruskal–Wallis
Checks whether any group differs.
Statistics
Why AutoDunn uses more than one test.
First check all groups. Then find which pairs differ.
First the whole set, then the pairs
Three simple stages explain the result.
Dunn pair comparisons
Compares groups two at a time.
Adjusted p-values
Adjustments reduce false alarms when many pairs are tested.
A small p-value does not tell the whole story.
A small p-value does not show how large or useful a difference is.
Also check: plots, data quality, group sizes, and real-world meaning.
test results + visuals + real-world context
→ better interpretation
→ better interpretation
Ways to adjust p-values
These methods reduce false positives when many pairs are tested.
| Approach | General idea | In simple terms |
|---|---|---|
| Bonferroni | Uses a stricter cutoff. | Simple and cautious. |
| Benjamini–Hochberg FDR | Limits false discoveries. | Useful for many comparisons. |
| Holm–Sidak | Adjusts cutoffs step by step. | Controls false positives. |