Visualizing Interactions
What you'll learn: How to create interaction plots that show how the relationship between a predictor and outcome changes across different groups or levels.
Why Visualize Interactions?
After creating interaction terms and testing their significance, the best way to understand what's happening is to see it. Numbers tell you if an interaction exists; plots show you what it looks like and how strong it is.
Think of it like this: you could describe a sunset with temperature readings and light wavelengths, or you could just show someone a photo. Interaction plots are that photo for your model.
What Interaction Plots Show
An interaction plot typically displays:
- The predictor variable on the x-axis
- The outcome variable on the y-axis
- Separate lines (or groups of points) for each level of the moderating variable
For categorical × continuous interactions: You'll see different slopes for each category. If the lines are parallel, there's no interaction. If they cross, diverge, or have noticeably different slopes, the interaction is present.
For continuous × continuous interactions: You might plot the relationship at different levels of the moderating variable (e.g., low/medium/high values), showing how the slope changes.
For categorical × categorical interactions: Bar charts grouped by one category and colored/positioned by another reveal how effects combine.
Reading the Story
When lines cross or fan out dramatically, you're seeing evidence that "it depends"—the effect of one variable truly depends on the value of another. Parallel lines mean each variable acts independently.
Key Takeaway: Interaction plots translate complex statistical relationships into visual stories, making it immediately clear how one predictor's effect changes across levels of another—essential for communicating and understanding your model.