Course contentsShow
← Data Science
Lesson 469 of 2,14512. ANOVA and Multiple ComparisonsFree lesson

Follow-Up Tests for Two-Way ANOVA

Performing simple effects tests and pairwise comparisons when interactions are significant.

Follow-Up Tests for Two-Way ANOVA

What you'll learn: How to investigate specific group differences after finding a significant interaction or main effect in two-way ANOVA.

Why Follow-Up Tests Matter

When your two-way ANOVA reveals a significant interaction effect, the F-test only tells you that something is happening between your factors—but not what exactly. Think of it like a smoke alarm: it alerts you to fire, but doesn't tell you which room is burning. Follow-up tests pinpoint the specific differences.

Simple Effects Tests

When an interaction is significant, you need to perform simple effects tests. These examine one factor at each level of the other factor separately.

For example, if you're studying how teaching method (Factor A) and class size (Factor B) affect test scores, and you find an interaction, you might ask:

  • "Does teaching method matter for small classes?"
  • "Does teaching method matter for large classes?"

You're essentially running separate one-way ANOVAs within slices of your data.

Pairwise Comparisons

After identifying where simple effects exist, use pairwise comparisons (like Tukey's HSD or Bonferroni) to determine which specific groups differ. This is the same approach you learned for one-way ANOVA post-hoc tests, but applied within the context of your interaction.

If there's no significant interaction but you have significant main effects, you can proceed directly to pairwise comparisons for each main effect independently—just like standard post-hoc testing.

The Workflow

  1. Check if interaction is significant
  2. If yes → simple effects tests at each level
  3. Follow with pairwise comparisons where needed
  4. If no → test main effects with standard post-hoc methods

Key Takeaway: When two-way ANOVA shows a significant interaction, use simple effects tests to examine one factor at specific levels of the other, then apply pairwise comparisons to identify exact group differences—this reveals the story behind the interaction.