When to Use the One-Sample Z-Test for Proportions
You'll learn when it's appropriate to test whether a population proportion matches a specific hypothesized value.
The Core Idea
Up until now, you've been testing means (averages) using t-tests and rank-based methods. But what if your data isn't about averages—what if it's about proportions? For example: "Is 60% of voters really in favor of this policy?" or "Does our website have a 10% click-through rate?"
The one-sample Z-test for proportions is designed specifically for these yes/no, success/failure, category questions where you want to compare an observed proportion from your sample against a hypothesized population proportion.
When to Use It
Use the one-sample Z-test for proportions when:
- Your data is categorical/binary: Each observation falls into one of two categories (success/failure, yes/no, clicked/didn't click)
- You have a single sample from one population
- You want to test against a specific proportion: You have a hypothesized value (e.g., H₀: p = 0.50)
- Sample size is large enough: Generally, you need both np ≥ 10 and n(1-p) ≥ 10, where n is sample size and p is the hypothesized proportion
- Observations are independent: Each observation doesn't influence others
Analogy
Think of flipping a coin. If someone claims a coin is fair (p = 0.50 heads), you flip it 100 times and get 65 heads. The one-sample Z-test for proportions helps you decide: is this coin really unfair, or did you just get lucky?
Key Takeaway: Use the one-sample Z-test for proportions when you have binary/categorical data from a single sample and want to test if the true population proportion equals a specific hypothesized value, assuming sufficient sample size and independence.