Relationship
Interaction plot
Shows response changes across one factor for different levels of another.
- Plot
- Relationship
- Interaction
What it answers
Does one factor's effect depend on another?
- Shape
- Response lines across factor levels
- What it shows
- Differential response and possible interaction.
Example
The question
Does the effect of a discount depend on the channel?
Average order value by discount and channel (interaction plot, EUR)
Illustrative data. It belongs to no company.
When to use it
Use it when
Exploring factorial or model-estimated interactions.
Avoid it when
Confusing visual nonparallelism with a confirmed statistical test.
How to read it
Compare response-line slopes across factor levels.
Note
Nonparallel lines do not establish statistical significance.
The data you need
- Layout
- Long table
- Fields
- FactorA: category; FactorB: category; Response: number
- One row is
- One raw response for a combination of two factors. What is the grain?
- Rules
- Retain replicates and aggregate with a stated statistic. Model-adjusted estimates require a fitted model.
- Optional
- Precomputed alternative: FactorA,FactorB,Estimate,Lower,Upper.
- Preparation
- Calculate response summaries or model-adjusted predictions for each factor combination and retain sample sizes/intervals.
- Units
- Use the measured variable's stated unit consistently. No denominator is implied unless the value is a rate or proportion.
Worked example
GroupA rises 10 to 20; GroupB15 to 18 from Low to High. Changes 10 and 3 differ by 7, suggesting an interaction to assess with a model.
| FactorA | FactorB | Response |
|---|---|---|
| Low | A | 10 |
| High | A | 20 |
| Low | B | 15 |
| High | B | 18 |