Interpreting Categorical × Continuous Interactions
What you'll learn: How to understand interaction terms between a dummy variable and a continuous predictor as different slopes for different groups.
The Core Idea
When you include an interaction between a categorical variable (encoded as dummy variables) and a continuous predictor, you're allowing each category to have its own slope for that continuous variable.
Without an interaction, all categories share the same slope—they just have different intercepts (parallel lines). With the interaction, the relationship between your continuous predictor and the outcome can be stronger, weaker, or even opposite depending on which category you're in.
How It Works
Imagine predicting salary from years of experience, with a dummy variable for whether someone has a graduate degree (0 = no, 1 = yes).
- Main effect of experience: the baseline slope (for the reference group, no degree)
- Main effect of degree: the intercept difference between groups
- Interaction term (degree × experience): how much steeper or flatter the experience slope is for people with a degree
If the interaction coefficient is positive, the degree group's slope is steeper—experience "pays off" more for them. If negative, experience matters less for the degree group.
Analogy
Think of two runners on different treadmills. The main effects tell you their starting positions and baseline speeds. The interaction tells you whether one treadmill accelerates faster than the other as time passes.
Key Takeaway: A categorical × continuous interaction lets each category have its own unique slope, revealing whether the continuous predictor's effect differs across groups.