Interpreting Coefficients Relative to Reference
What You'll Learn
You'll learn why dummy variable coefficients don't represent absolute values, but rather differences from the reference group.
The Core Idea
When you use dummy variables in regression, each coefficient tells you how much higher or lower that category is compared to the reference category—not its absolute level.
Think of it like comparing salaries at different job titles. If "Junior Developer" is your reference category earning 30,000, that doesn't mean seniors earn 30,000 more than juniors**—so $90,000 in absolute terms.
How It Works
When you have a categorical predictor (like city: Boston, NYC, LA) and choose Boston as your reference:
- Boston's coefficient: doesn't appear (it's built into the intercept)
- NYC's coefficient (say, +15): means NYC is 15 units higher than Boston
- LA's coefficient (say, -5): means LA is 5 units lower than Boston
The intercept represents the baseline value for the reference group (Boston). All other coefficients are adjustments from that baseline.
Why This Matters
This relative interpretation is crucial for:
- Correctly explaining results: "NYC homes cost 15k")
- Statistical testing: you're testing whether categories differ from the reference, not whether they differ from zero
- Choosing references wisely: pick a meaningful baseline for comparison
Key Takeaway: Dummy variable coefficients represent differences from the reference category, not standalone effects—always interpret them as "compared to [reference group]."