What you'll learn
This guide covers inference procedures for categorical data involving proportions, a fundamental topic in AP Statistics. You will master the techniques for constructing confidence intervals and conducting hypothesis tests for both one-sample and two-sample proportion scenarios. These methods enable you to draw valid conclusions about population proportions from sample data.
Key terms and definitions
Sample proportion (p̂) — the fraction of individuals in a sample with a particular characteristic, calculated as x/n where x is the count of successes and n is the sample size.
Standard error — the estimated standard deviation of the sampling distribution of a statistic; for a proportion, SE = √[p̂(1-p̂)/n].
Sampling distribution — the probability distribution of a statistic obtained from all possible samples of a given size from a population.
One-proportion z-test — a hypothesis test procedure used to determine whether a population proportion differs from a hypothesized value.
Two-proportion z-test — a hypothesis test procedure used to compare proportions from two independent populations or treatments.
Confidence interval — a range of plausible values for a population parameter, constructed from sample data with a specified confidence level.
Null hypothesis (H₀) — a statement of no effect or no difference, typically claiming a parameter equals a specific value or that two parameters are equal.
Type I error — rejecting the null hypothesis when it is actually true; the probability of this error is denoted by α (significance level).
Core concepts
Conditions for inference about proportions
Before conducting any inference procedure for proportions, you must verify three essential conditions:
Random condition: The data must come from a random sample or randomized experiment. State explicitly: "The problem states this is a random sample" or "Subjects were randomly assigned to treatments."
Normal/Large Counts condition: The sampling distribution must be approximately normal. For confidence intervals, check that np̂ ≥ 10 and n(1-p̂) ≥ 10. For significance tests, check that np₀ ≥ 10 and n(1-p₀) ≥ 10, where p₀ is the null hypothesis proportion.
Independence/10% condition: Individual observations must be independent. When sampling without replacement, the sample size should not exceed 10% of the population size (n ≤ N/10).
For two-proportion procedures, verify these conditions for both samples separately.
One-proportion z-interval
A one-proportion z-interval estimates a population proportion p with a specified confidence level (commonly 90%, 95%, or 99%).
Formula: p̂ ± z* √[p̂(1-p̂)/n]
Where:
- p̂ is the sample proportion
- z* is the critical value from the standard normal distribution
- n is the sample size
Common critical values:
- 90% confidence: z* = 1.645
- 95% confidence: z* = 1.96
- 99% confidence: z* = 2.576
Interpretation template: "We are [confidence level]% confident that the true proportion of [context] is between [lower bound] and [upper bound]."
Always interpret in context of the problem, never stating "we are 95% confident that p̂ is between..." The interval estimates the parameter p, not the statistic p̂.
One-proportion z-test
This test evaluates whether sample data provides convincing evidence that a population proportion differs from a hypothesized value.
Test statistic: z = (p̂ - p₀) / √[p₀(1-p₀)/n]
Note that the standard error uses p₀ (the null hypothesis value), not p̂. This represents the variability we would expect if H₀ were true.
Steps for conducting the test:
State hypotheses:
- H₀: p = p₀ (or p ≤ p₀ or p ≥ p₀)
- Hₐ: p ≠ p₀ (two-sided) or p > p₀ or p < p₀ (one-sided)
Verify conditions: Random, Normal (large counts), and Independence
Calculate test statistic: Use the formula above
Find P-value:
- Two-sided: P(|Z| ≥ |z|) = 2 × P(Z ≥ |z|)
- One-sided (>): P(Z ≥ z)
- One-sided (<): P(Z ≤ z)
Make decision: If P-value < α, reject H₀; otherwise, fail to reject H₀
State conclusion in context: Link the statistical decision to the real-world question
Two-proportion z-interval
This procedure estimates the difference between two population proportions, p₁ - p₂.
Formula: (p̂₁ - p̂₂) ± z* √[p̂₁(1-p̂₁)/n₁ + p̂₂(1-p̂₂)/n₂]
The standard error combines variability from both samples. Each sample proportion contributes its own variance term.
Interpretation: "We are [confidence level]% confident that the true difference in proportions of [group 1 context] and [group 2 context] is between [lower bound] and [upper bound]."
If the interval contains 0, this suggests no convincing evidence of a difference between the populations. If the interval is entirely positive, p₁ is likely greater than p₂; if entirely negative, p₁ is likely less than p₂.
Two-proportion z-test
This test determines whether two population proportions differ significantly.
Pooled proportion: p̂_c = (x₁ + x₂)/(n₁ + n₂)
Under the null hypothesis of no difference (p₁ = p₂), we combine the samples to get a better estimate of the common proportion.
Test statistic: z = (p̂₁ - p̂₂) / √[p̂_c(1-p̂_c)(1/n₁ + 1/n₂)]
Standard hypothesis structure:
- H₀: p₁ = p₂ (equivalently: p₁ - p₂ = 0)
- Hₐ: p₁ ≠ p₂ (or p₁ > p₂ or p₁ < p₂)
The procedure follows the same six steps as the one-proportion test, but verify conditions for both samples independently.
Choosing between procedures
One-sample vs two-sample: Use one-sample procedures when comparing a single population proportion to a known value. Use two-sample procedures when comparing proportions from two independent groups.
Confidence interval vs hypothesis test: Construct a confidence interval when estimating a parameter or range of plausible values. Conduct a hypothesis test when evaluating a specific claim about a parameter.
One-sided vs two-sided tests: Use a one-sided alternative when you have a directional research question (testing if a proportion is greater than or less than a value). Use a two-sided alternative when simply testing for a difference in either direction.
Worked examples
Example 1: One-proportion z-interval
A random sample of 250 students at a Caribbean secondary school found that 165 students have internet access at home. Construct and interpret a 95% confidence interval for the proportion of all students at the school with home internet access.
Solution:
Conditions:
- Random: The problem states a random sample was selected ✓
- Normal: p̂ = 165/250 = 0.66; np̂ = 250(0.66) = 165 ≥ 10 and n(1-p̂) = 250(0.34) = 85 ≥ 10 ✓
- Independence: 250 students is certainly less than 10% of all students at the school ✓
Calculations:
- p̂ = 165/250 = 0.66
- For 95% confidence, z* = 1.96
- SE = √[0.66(0.34)/250] = √0.000898 = 0.0300
- Margin of error = 1.96(0.0300) = 0.0588
- Interval: 0.66 ± 0.0588 = (0.601, 0.719)
Interpretation: We are 95% confident that the true proportion of all students at this school with internet access at home is between 0.601 and 0.719 (or between 60.1% and 71.9%).
Example 2: One-proportion z-test
A manufacturer claims that 90% of their products pass quality inspection. A consumer group randomly selects 200 products and finds that 170 pass inspection. At the α = 0.05 significance level, is there convincing evidence that the true pass rate is less than the manufacturer's claim?
Solution:
Hypotheses:
- H₀: p = 0.90
- Hₐ: p < 0.90 (one-sided test)
Conditions:
- Random: Random selection stated ✓
- Normal: np₀ = 200(0.90) = 180 ≥ 10 and n(1-p₀) = 200(0.10) = 20 ≥ 10 ✓
- Independence: 200 is less than 10% of all products manufactured ✓
Calculations:
- p̂ = 170/200 = 0.85
- z = (0.85 - 0.90) / √[0.90(0.10)/200]
- z = -0.05 / √0.00045 = -0.05 / 0.0212 = -2.36
- P-value = P(Z ≤ -2.36) = 0.0091
Decision: Since P-value (0.0091) < α (0.05), we reject H₀.
Conclusion: There is convincing evidence that the true proportion of products passing quality inspection is less than 90%.
Example 3: Two-proportion z-test
Researchers conducted a randomized experiment to test a new revision app. Of 120 students randomly assigned to use the app, 84 passed their exam. Of 100 students randomly assigned to traditional revision methods, 60 passed. Is there convincing evidence at α = 0.01 that the app increases pass rates?
Solution:
Hypotheses:
- H₀: p_app = p_trad (or p_app - p_trad = 0)
- Hₐ: p_app > p_trad (one-sided)
Conditions:
- Random: Random assignment to treatments ✓
- Normal: App group: 84 ≥ 10, 36 ≥ 10; Traditional: 60 ≥ 10, 40 ≥ 10 ✓
- Independence: Random assignment ensures independence; both sample sizes < 10% of all possible students ✓
Calculations:
- p̂_app = 84/120 = 0.70
- p̂_trad = 60/100 = 0.60
- p̂_c = (84 + 60)/(120 + 100) = 144/220 = 0.6545
- z = (0.70 - 0.60) / √[0.6545(0.3455)(1/120 + 1/100)]
- z = 0.10 / √[0.2263(0.0183)] = 0.10 / √0.00414 = 0.10 / 0.0643 = 1.56
- P-value = P(Z ≥ 1.56) = 0.0594
Decision: Since P-value (0.0594) > α (0.01), we fail to reject H₀.
Conclusion: There is not convincing evidence at the 0.01 significance level that the revision app increases pass rates compared to traditional methods.
Common mistakes and how to avoid them
Using p̂ in the standard error for hypothesis tests: In significance tests, always use p₀ (the null hypothesis value) in the standard error formula, not the sample proportion. Save p̂ for confidence intervals.
Incorrect condition checking: Check np and n(1-p) values using the appropriate proportion—p̂ for intervals, p₀ for tests, both sample proportions for two-proportion intervals, and the pooled proportion for two-proportion tests. Write out the actual calculations.
Misinterpreting confidence intervals: Never say "95% of the data falls in this interval" or "there is a 95% probability that p is in this interval." The correct interpretation involves confidence about where the true parameter lies.
Confusing failure to reject with accepting H₀: When P-value > α, state "we fail to reject H₀" or "there is not convincing evidence for Hₐ." Never conclude that H₀ is true or proven.
Forgetting context in conclusions: Always relate statistical conclusions back to the specific scenario. Don't just say "reject H₀"—explain what this means for pass rates, internet access, or whatever the context involves.
Mixing up one-sided and two-sided P-values: For one-sided tests, use the area in one tail only. For two-sided tests, double the tail probability. Match the alternative hypothesis direction carefully.
Exam technique for "Inference for Categorical Data: Proportions"
Show all condition checks explicitly: AP Statistics examiners award marks for properly stating and verifying conditions. Write "Random:", "Normal:", and "Independence:" as separate lines with explanations, even when the problem seems obvious.
Use the four-step process for inference problems: (1) State parameter and hypotheses or confidence level, (2) Check conditions, (3) Calculate statistic and P-value or interval, (4) Conclude in context. This structure earns maximum marks and keeps your work organized.
Match your conclusion to the question asked: If the question asks "Is there evidence that...?", your conclusion must address that specific claim. If asked to construct an interval, include a complete interpretation statement even if not explicitly requested.
Know your formulas: The AP Statistics formula sheet provides formulas for test statistics and confidence intervals, but you must know when to apply each one and how to identify the components correctly.
Quick revision summary
Inference for proportions requires verifying Random, Normal (large counts), and Independence conditions before proceeding. One-sample procedures compare a population proportion to a hypothesized value; two-sample procedures compare proportions between independent groups. Use p̂ in standard errors for confidence intervals but p₀ for hypothesis tests. The pooled proportion combines both samples in two-proportion tests. Always interpret confidence intervals and hypothesis test conclusions in the context of the problem, relating statistical results to real-world questions about the populations of interest.