Course contentsShow
← Data Science
Lesson 341 of 2,1459. Hypothesis Testing FoundationsFree lesson

Effect Size and Power

Larger true effects are easier to detect; power increases with the magnitude of the effect.

Effect Size and Power

You'll learn: How the magnitude of a real effect in your population influences your test's ability to detect it.

The Core Idea

Imagine you're trying to spot a friend in a crowd. If your friend is wearing a bright neon jacket (large effect), you'll spot them easily. If they're wearing all black in a sea of black-clad people (small effect), it's much harder to pick them out—even though they're there.

Statistical power works the same way. Effect size measures how big the real difference or relationship is in your population. When the true effect is large, your hypothesis test has an easier time detecting it—meaning your statistical power increases.

For example:

  • A new drug that reduces symptoms by 50% (large effect) is easier to detect in a clinical trial than one that reduces symptoms by 5% (small effect)
  • A 20-point difference in test scores (large effect) requires a smaller sample to detect than a 2-point difference (small effect)

Why This Matters

When you're designing a study, knowing the expected effect size helps you understand how much power you'll have. If you're hunting for a subtle effect (like a 1% improvement in conversion rate), you'll need either:

  • A larger sample size, or
  • Accept lower power (higher risk of missing the effect)

If the effect is large, even a modest sample might give you excellent power to detect it.

Key Takeaway: Larger true effects are easier to detect and automatically increase your test's statistical power—like trying to spot a lighthouse versus a candle from a distance.