Concluding Tests for a
Population Proportion
Master the final step of significance testing: comparing p-values to significance levels, making reject or fail-to-reject decisions, and writing conclusions in context.
What Is a Hypothesis Test?
A hypothesis test for a population proportion \( p \) asks whether sample data provide convincing evidence against a claimed value \( p_0 \). We frame the question with two competing statements:
Understanding the p-value
The p-value is the probability of obtaining a test statistic at least as extreme as the observed one, assuming \( H_0 \) is true. It measures how surprising the sample result is under the null hypothesis.
The Significance Level \( \alpha \)
The significance level \( \alpha \) is the threshold we set before collecting data. It is the maximum p-value for which we will reject \( H_0 \). Common choices:
The Decision Rule
The entire conclusion hinges on one comparison:
Reject vs. Fail to Reject
| Reject \( H_0 \) | Fail to Reject \( H_0 \) | |
|---|---|---|
| Condition | p-value \( \leq \alpha \) | p-value \( > \alpha \) |
| Evidence | Convincing evidence for \( H_a \) | Not convincing evidence for \( H_a \) |
| Language | "We have convincing evidence that..." | "We do not have convincing evidence that..." |
| Error Risk | Type I error possible | Type II error possible |
Writing Conclusions in Context
A complete AP conclusion requires three components. Missing any one can cost you points:
Step-by-Step: How to Conclude
Worked Example: Coin Fairness
A student flips a coin 200 times and gets 115 heads. She tests \( H_0\!: p = 0.50 \) vs. \( H_a\!: p \neq 0.50 \) at \( \alpha = 0.05 \). The computed p-value is 0.034.
Worked Example: Voter Support
A pollster claims more than 60% of voters support a policy. In a random sample of 400 voters, 252 support it. Testing \( H_0\!: p = 0.60 \) vs. \( H_a\!: p > 0.60 \) at \( \alpha = 0.05 \), the p-value is 0.110.
Worked Example: Defective Parts
A factory claims its defective rate is below 3%. A quality inspector tests \( H_0\!: p = 0.03 \) vs. \( H_a\!: p < 0.03 \) with \( n = 500 \) parts and finds 7 defective. The test yields a p-value of 0.018 at \( \alpha = 0.05 \).
Type I and Type II Errors
Every decision carries a risk of error. Understanding these errors helps you interpret your conclusion's reliability.
One-Sided vs. Two-Sided Tests
| One-Sided | Two-Sided | |
|---|---|---|
| \( H_a \) | \( p > p_0 \) or \( p < p_0 \) | \( p \neq p_0 \) |
| Tail(s) | One tail | Both tails |
| Conclusion wording | "...greater than" or "...less than" | "...different from" |
| p-value size | Smaller (easier to reject) | Larger (harder to reject) |
Common Mistakes
Summary: Conclusion Checklist
| Step | What to Do | Example Phrase |
|---|---|---|
| 1. State p-value and \( \alpha \) | Quote both numerical values | "p-value = 0.034, \( \alpha = 0.05 \)" |
| 2. Compare | Use \( \leq \) or \( > \) | "Since 0.034 \( \leq \) 0.05..." |
| 3. Decision | Reject or fail to reject \( H_0 \) | "...we reject \( H_0 \)." |
| 4. Context | Reference the population and claim | "There is convincing evidence that..." |
Worked Example: Full Hypothesis Test with Conclusion
A hospital administrator claims that fewer than 15% of patients discharged from the emergency department are readmitted within 30 days. A random sample of 200 discharged patients reveals that 22 were readmitted. At the α = 0.05 significance level, is there convincing evidence that the readmission rate is less than 15%? One-Proportion z-Test for Hospital Readmission Rate Step 1 — State the Hypotheses Define p as the true proportion of all emergency department patients at this hospital who are readmitted within 30 days. The null hypothesis is H₀: p = 0.15 and the alternative hypothesis is Hₐ: p < 0.15. This is a one-sided (left-tailed) test because the administrator's claim suggests the rate is below 15%. Step 2 — Check Conditions Random: The sample is described as a random sample of 200 discharged patients. ✓ Independence: Assuming the hospital discharges far more than 10 × 200 = 2,000 patients, the 10% condition is met. ✓ Large Counts: np₀ = 200(0.15) = 30 ≥ 10 and n(1 − p₀) = 200(0.85) = 170 ≥ 10, so the normal approximation is appropriate. ✓ Step 3 — Compute the Test Statistic The sample proportion is p̂ = 22/200 = 0.11. The standard error under H₀ is √(0.15 × 0.85 / 200) = √(0.1275 / 200) = √0.0006375 ≈ 0.02525. Therefore, z = (0.11 − 0.15) / 0.02525 = −0.04 / 0.02525 ≈ −1.584. z ≈ −1.584 Step 4 — Find the P-Value Because this is a left-tailed test (Hₐ: p < 0.15), the p-value is P(Z ≤ −1.584). Using a standard normal table or calculator: p-value ≈ 0.0566. p-value ≈ 0.0566 Step 5 — State the Conclusion Because the p-value (0.0566) is greater than α = 0.05, we fail to reject H₀. There is not convincing evidence that the true proportion of emergency department patients at this hospital who are readmitted within 30 days is less than 0.15. Fail to reject H₀ — not convincing evidence that p < 0.15 Why Step 5 Matters Most Notice that the conclusion does not say 'we accept that p = 0.15' or 'the readmission rate is 15%.' It only says we lack convincing evidence that it is lower. The data are consistent with p = 0.15, but also consistent with many other values. This nuanced language is essential and reflects the Neyman–Pearson logic: the test is designed to control the probability of wrongly rejecting H₀, not to confirm it.


Question 1 of 5
A researcher tests \( H_0\!: p = 0.40 \) vs. \( H_a\!: p > 0.40 \) at \( \alpha = 0.05 \) and obtains a p-value of 0.023. What is the correct conclusion?
Question 2 of 5
A test of \( H_0\!: p = 0.75 \) vs. \( H_a\!: p \neq 0.75 \) yields a p-value of 0.08. At \( \alpha = 0.05 \), what should you conclude?
Question 3 of 5
Which of the following is the best conclusion for a test where the p-value is 0.002 and \( \alpha = 0.01 \)?
Question 4 of 5
A student concludes: "We reject \( H_0 \). The true proportion is definitely not 0.50." What is wrong with this conclusion?
Question 5 of 5
When we fail to reject \( H_0 \), which type of error is possible?