Learn
Interactive·

Multiple Comparisons

A significant ANOVA says that at least one mean differs. How do we find the groups responsible?

An ANOVA comparing three delivery methods produced a statistically significant result.

Delivery methodMean delivery time
Standard delivery42 hours
Next-day delivery35 hours
Local hub delivery28 hours

We know that the equal-means model does not fit well, but ANOVA does not identify the groups that differ.

Key question

How can several groups be compared without allowing the false-positive rate to grow unchecked?

1

Concept

Why do repeated p-values create a problem?

More comparisons create more opportunities for error

One test at the 0.05 level has a 5% false-positive rate under the null. Across several tests, the probability that at least one crosses the threshold by chance becomes much larger.

Number of comparisonsProbability of at least one false positive
15%
3About 14%
6About 26%
10About 40%

Familywise error rate

The probability of at least one false positive across a family of comparisons is called the familywise error rate.

2

Why It Matters

Which post hoc method should you use?

The right method depends on the comparison plan

Consider whether every pair matters, whether one group is a control, and whether group variances are similar.

All pairs

Tukey HSD

A common choice when every pair of means matters

Unequal variances

Games–Howell

Useful when variances or sample sizes differ

Selected comparisons

Bonferroni

Simple and conservative for a small comparison set

One control

Dunnett

Compares several groups with one reference group

A common choice

After a standard one-way ANOVA, Tukey HSD is common for all pairwise comparisons. When variances or sample sizes differ substantially, Games–Howell is usually safer.

3

How It Works

What does adjustment change?

Adjustment does not change the observed mean difference

The difference remains the same. The adjusted p-value, confidence interval, and evidence threshold change.

Adjustment can change the conclusion without changing the difference

When all three pairs are tested, raw p-values should not be read in isolation.

ComparisonMean differenceRaw pAdjusted pConclusion
Standard − next-day7 hours0.0480.144Not significant
Standard − local hub14 hours< 0.001< 0.001Significant
Next-day − local hub7 hours0.0480.144Not significant

The tradeoff

Stronger control of false positives can make real differences harder to detect. Match the method to the planned comparisons.

4

Example

How should the results be read?

Read the overall test and follow-up comparisons together

Post hoc results

Tukey HSD · familywise error rate 5%

3 comparisons
Standard vs next-dayDifference 7 hNot significant
Standard vs local hubDifference 14 hSignificant
Next-day vs local hubDifference 7 hNot significant
  • Check the overall ANOVA first.
  • Choose a follow-up method that matches the research question.
  • Read adjusted p-values and simultaneous confidence intervals.
  • Interpret mean differences and effect sizes as well.

Reporting example

Post hoc comparisons showed that local hub delivery was 14 hours faster than standard delivery (adjusted 95% CI [5.5, 22.5], adjusted p < .001). The 7-hour difference from next-day delivery was not significant after adjustment (adjusted 95% CI [−1.5, 15.5], adjusted p = .144).

5

Interactive

Change the number of comparisons and adjustment

Change the number of comparisons and the adjustment

The significance threshold, adjusted p-value, and false-positive risk update together.

Comparisons

3

Raw p-value

0.031

Adjusted p-value

0.093

Conclusion

Not significant

Unadjusted familywise error rate14.3%

Per-comparison threshold: p < 0.0167

Key lesson

Adjustment is not a way to make p-values smaller. It is a way to reduce overconfidence created by repeated comparisons.

Key takeaways

  • ANOVA shows only that at least one population mean differs.
  • Post hoc tests identify the relevant group differences.
  • False-positive risk grows as more comparisons are performed.
  • Tukey HSD is commonly used for all pairwise comparisons.
  • Games–Howell is useful when variances or sample sizes differ.
  • Adjusted p-values, intervals, and mean differences should be interpreted together.

Post hoc tests answer “which groups differ?” while accounting for the fact that several comparisons were examined.

Now compare groups with a real datasetOpen in Lab

Go Deeper

Test Assumptions