Interpreting Categorical × Categorical Interactions
What you'll learn: How to understand what happens when the effect of one categorical variable depends on the level of another categorical variable.
The Core Idea
When you interact two categorical variables, you're asking: "Does the difference between groups in one variable change depending on which group you're in for the other variable?"
Imagine studying the effect of Treatment Type (Drug A vs Drug B) and Gender (Male vs Female) on recovery time. Without an interaction, you'd assume Drug A beats Drug B by the same amount for both males and females. But an interaction allows Drug A to work better than Drug B for males while Drug B works better for females—the treatment effect depends on gender.
How It Works with Dummy Variables
Remember that categorical variables become dummy variables. If you have two binary categoricals:
- Treatment: Drug A (reference) vs Drug B
- Gender: Male (reference) vs Female
Your model includes:
- A dummy for Drug B
- A dummy for Female
- An interaction term: Drug B × Female (equals 1 only when both are true)
The interaction coefficient tells you: "How much does the Drug B effect change when you're female compared to male?"
Reading the Combinations
The model creates a separate effect for each combination of categories. With two binary variables, you get four groups: Drug A + Male (baseline), Drug A + Female, Drug B + Male, Drug B + Female. The interaction lets each combination have its own unique mean, not just additive effects.
Key Takeaway: Categorical × categorical interactions allow group differences to vary across combinations, capturing when the effect of one grouping depends on another grouping—essential for modeling complex real-world scenarios where effects aren't uniform.