Setting Up Hypotheses for Proportions
You'll learn how to formulate null and alternative hypotheses when testing claims about population proportions.
The Core Idea
When testing proportions—like whether a new website design converts more visitors, or if a medication's side-effect rate differs from the advertised value—you need to translate your research question into statistical hypotheses. Just as you did with means in earlier t-tests, you'll define a null hypothesis (H₀) representing "no effect" or the status quo, and an alternative hypothesis (H₁) representing what you suspect or want to detect.
The key difference: instead of testing μ (population mean), you're testing p (population proportion).
How to Formulate Proportion Hypotheses
The Null Hypothesis (H₀)
This states that the population proportion equals a specific value, typically a known or claimed proportion:
- H₀: p = 0.50 (the coin is fair)
- H₀: p = 0.05 (the side-effect rate is 5% as claimed)
- H₀: p = 0.30 (30% of voters support the policy)
The Alternative Hypothesis (H₁)
This reflects your research question and can be two-sided or one-sided:
Two-sided (testing for any difference):
- H₁: p ≠ 0.50
One-sided (testing for a specific direction):
- H₁: p > 0.30 (support has increased)
- H₁: p < 0.05 (side effects are less common than claimed)
Real-World Example
A pharmaceutical company claims their drug causes nausea in 10% of patients. You suspect it's actually higher.
- H₀: p = 0.10 (the claim is correct)
- H₁: p > 0.10 (nausea rate exceeds the claim)
Key Takeaway: When testing proportions, set H₀ as "p equals the claimed value" and H₁ based on whether you're testing for any difference (two-sided) or a specific direction (one-sided).