Interpreting Logistic Regression Coefficients
What you'll learn: How to translate logistic regression coefficients from log-odds ratios into meaningful odds ratios that you can actually explain.
The Core Challenge
When you fit a logistic regression model, the coefficients you get represent log-odds ratios—not probabilities or simple effects. This makes them tricky to interpret at first glance. A coefficient of 0.5 doesn't mean "the probability increases by 0.5" or "a 50% increase." Instead, it means something more subtle.
From Log-Odds to Odds Ratios
The coefficient β tells you how the log-odds of your outcome change per one-unit increase in your predictor (holding others constant). To make this interpretable:
Exponentiate the coefficient: exp(β)
This gives you the odds ratio (OR).
What Does the Odds Ratio Mean?
- OR = 1: No effect (predictor doesn't change the odds)
- OR > 1: Predictor increases the odds of the outcome
- OR < 1: Predictor decreases the odds of the outcome
Example: If β = 0.693, then exp(0.693) ≈ 2.0. This means each one-unit increase in your predictor doubles the odds of the outcome occurring.
Another example: If β = -0.693, then exp(-0.693) ≈ 0.5. Each one-unit increase in the predictor halves the odds.
Practical Interpretation
Think of odds ratios as multipliers. An OR of 1.5 means the odds are 1.5 times higher (a 50% increase). An OR of 0.8 means the odds are 0.8 times as high (a 20% decrease).
Key Takeaway: Logistic regression coefficients are log-odds ratios; exponentiate them to get odds ratios, which tell you the multiplicative change in odds for each unit increase in your predictor.