Course contentsShow
← Data Science
Lesson 688 of 2,14518. Generalized Linear ModelsFree lesson

Effect Size and Practical Significance

Interpreting the magnitude of odds ratios, distinguishing statistical from practical significance in logistic models.

Effect Size and Practical Significance

You'll learn how to judge whether a statistically significant logistic regression result actually matters in the real world.

The Core Idea

In logistic regression, you've learned to test whether coefficients differ from zero (statistical significance) and interpret odds ratios. But statistical significance doesn't mean practical importance. An odds ratio of 1.02 might be "statistically significant" with enough data, but does a 2% increase in odds really matter?

Effect size in logistic models refers to the magnitude of the odds ratio. Practical significance asks: "Is this effect large enough to care about?"

Interpreting Odds Ratio Magnitude

  • OR ≈ 1.0–1.2: Small effect (20% change in odds or less)
  • OR ≈ 1.5–2.0: Moderate effect
  • OR ≥ 3.0: Large effect

The same logic applies below 1.0 (OR = 0.5 means a 50% decrease in odds—a large protective effect).

Statistical vs Practical Significance

Imagine a medical study finds that a new drug increases recovery odds by OR = 1.05 (p < 0.001). With 100,000 patients, this tiny effect is statistically significant (the p-value is tiny), but a 5% increase may not justify the drug's cost or side effects—it lacks practical significance.

Conversely, a smaller study might find OR = 2.5 (p = 0.08). Not statistically significant at α = 0.05, but the size of the effect suggests clinical importance worth investigating further.

The Takeaway

Always report both the odds ratio (effect size) and the confidence interval alongside the p-value. A statistically significant result with OR ≈ 1.01 is scientifically weak; a large OR with a wide confidence interval crossing 1.0 might warrant further study despite non-significance.

Key Takeaway: Statistical significance tells you an effect is detectable; practical significance tells you if it's meaningful—in logistic regression, judge both the p-value and the magnitude of the odds ratio.