AP Statistics Unit 6 ReviewProportions

AP Statistics Unit 6 is where you finally do inference, drawing conclusions about an entire population from one sample of categorical data. The single biggest idea is that a sample proportion p̂ varies from sample to sample, so any conclusion about the true population proportion p has to be wrapped in uncertainty, either as a confidence interval (a range of plausible values) or a significance test (a yes-or-no decision with a known error rate). Unit 6 makes up 12-15% of the AP exam, and the four-step inference structure you learn here repeats in Units 7, 8, and 9.

What this unit covers

Estimating a proportion with a confidence interval

• A one-sample z-interval for a proportion answers "what is p, roughly?" The recipe is point estimate plus or minus margin of error, which becomes $\hat{p} \pm z^* \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$.
• Before you build the interval, you check conditions. Random sampling (or random assignment) gives you independence, the 10% condition ($n \leq 0.10N$) handles sampling without replacement, and the Large Counts condition ($n\hat{p} \geq 10$ and $n(1-\hat{p}) \geq 10$) justifies using a normal model.
• Interpretation is its own skill with its own template. "We are 95% confident that the interval from ___ to ___ captures the true proportion of [population] who [context]." Confidence is in the method, not in any single interval. In repeated sampling, about 95% of intervals built this way capture p.
• You also need the relationships. Bigger n means a narrower interval (width is proportional to $1/\sqrt{n}$), higher confidence level means a wider interval, and the width is exactly twice the margin of error.

Testing a claim about one proportion

• A one-sample z-test answers "is the true proportion really p₀, or is the sample evidence against that?" The null hypothesis H₀: p = p₀ is assumed true until evidence says otherwise; the alternative (Hₐ: p < p₀, p > p₀, or p ≠ p₀) is the claim you're gathering evidence for.
• The test statistic standardizes how far your sample fell from the null value: $z = \frac{\hat{p} - p_0}{\sqrt{\frac{p_0(1-p_0)}{n}}}$. Note that conditions and standard error use p₀ here, not p̂, because the whole calculation assumes H₀ is true.
• The p-value is the probability of getting a sample result as extreme or more extreme than yours, assuming H₀ is true. A tiny p-value means your data would be weird in a world where the null holds, which is evidence against the null.
• The decision rule is mechanical. If the p-value is at most α, reject H₀ and conclude there is convincing evidence for Hₐ. If the p-value is greater than α, fail to reject H₀. You never "accept" the null and you never "prove" anything.

Type I and Type II errors and power

• A Type I error rejects a true null hypothesis (a false positive). Its probability is exactly α, the significance level you chose.
• A Type II error fails to reject a false null hypothesis (a false negative). Power, the probability of correctly rejecting a false null, equals 1 minus P(Type II error).
• Power goes up when sample size increases, when α increases, when the standard error shrinks, or when the true parameter is farther from the null value.
• Which error is worse depends on context. If a Type I error has serious consequences (convicting an innocent person, recalling a safe product), you pick a smaller α. The exam loves asking you to describe both errors in context and weigh their consequences.

Comparing two proportions

• Everything doubles. A two-sample z-interval estimates p₁ − p₂ with $(\hat{p}_1 - \hat{p}_2) \pm z^* \sqrt{\frac{\hat{p}_1(1-\hat{p}_1)}{n_1} + \frac{\hat{p}_2(1-\hat{p}_2)}{n_2}}$, and a two-sample z-test checks H₀: p₁ = p₂.
• Conditions now apply to both groups. Two independent random samples (or a randomized experiment), both 10% conditions, and Large Counts for all four counts (successes and failures in each sample).
• The two-sample test has one twist. Because H₀ says the two proportions are equal, you pool the samples into a combined proportion $\hat{p}_c = \frac{n_1\hat{p}_1 + n_2\hat{p}_2}{n_1 + n_2}$ and use it in the standard error. Pooling happens only in the test, never in the interval.
• If a confidence interval for p₁ − p₂ contains 0, the data are consistent with no difference between the groups. If the whole interval sits above or below 0, you have evidence of a real difference, and the interval tells you how big.

Unit 6, Proportions at a glance

Why Unit 6, Proportions matters in AP Stats

Unit 6 is the payoff for everything in the first half of the course. The course's enduring themes (variation creates uncertainty, intervals account for that uncertainty, and significance tests let you make decisions anyway) all become concrete procedures here for the first time.

• This is the first unit where you go from describing a sample to making a defensible claim about a population, which is the entire point of inferential statistics.
• The four-step structure you build here (name the procedure, check conditions, calculate, conclude in context) is the template for every inference problem for the rest of the course.
• Type I and Type II errors formalize a big course theme, that statistical conclusions can be wrong in predictable ways, and that you get to choose which kind of wrongness you can tolerate.
• Proportion inference is everywhere in real life, from polling margins of error to clinical trial comparisons, so this unit also gives you the most directly applicable skill in the course.

How this unit connects across the course

• Sampling distributions (Unit 5) are the engine under the hood. The standard error $\sqrt{p(1-p)/n}$ and the Large Counts condition come straight from the sampling distribution of p̂. Unit 6 just adds decision-making on top.
• Whether your conclusion generalizes depends on data collection (Unit 3). Random sampling lets you infer about a population; random assignment in an experiment lets you infer cause and effect. Condition checks in Unit 6 explicitly reference these designs.
• The entire interval-and-test framework repeats with means in Unit 7, just swapping z for t and proportions for averages. If you really understand Unit 6, Unit 7 feels like review with a new distribution.
• When categorical data has more than two categories or two variables, the z-procedures break down, and chi-square tests (Unit 8) take over. Unit 8 is the generalization of Unit 6's two-sample test.

Key formulas and procedures

• $SE_{\hat{p}} = \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$, the estimated standard deviation of the sample proportion, used for confidence intervals.
• $\hat{p} \pm z^* \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$, the one-sample z-interval; the part after the ± is the margin of error.
• $z = \frac{\hat{p} - p_0}{\sqrt{p_0(1-p_0)/n}}$, the one-sample z-test statistic; uses the null value p₀ because the test assumes H₀ is true.
• $(\hat{p}_1 - \hat{p}_2) \pm z^* \sqrt{\frac{\hat{p}_1(1-\hat{p}_1)}{n_1} + \frac{\hat{p}_2(1-\hat{p}_2)}{n_2}}$, the two-sample z-interval for a difference of proportions.
• $\hat{p}_c = \frac{n_1\hat{p}_1 + n_2\hat{p}_2}{n_1 + n_2}$, the pooled (combined) proportion used only in the two-sample test.
• $z = \frac{(\hat{p}_1 - \hat{p}_2) - 0}{\sqrt{\hat{p}_c(1-\hat{p}_c)}\sqrt{\frac{1}{n_1} + \frac{1}{n_2}}}$, the two-sample z-test statistic.
• Decision rule: if p-value ≤ α, reject H₀; if p-value > α, fail to reject H₀. Always state the conclusion in the context of the problem.
• Condition checks for every procedure: random data collection, the 10% condition when sampling without replacement, and Large Counts (expected successes and failures both at least 10).
• The test statistic formulas are not printed on the exam formula sheet, but you can rebuild them from the general pattern (statistic minus parameter, divided by standard deviation of the statistic) plus the standard error formulas that are provided.

Unit 6, Proportions on the AP exam

Unit 6 carries 12-15% of the exam, and proportion inference is a staple of the free-response section. A full inference FRQ typically asks you to name the appropriate procedure (one-sample z-interval, two-sample z-test, and so on), verify conditions with actual numbers, compute the interval or test statistic and p-value, and write a conclusion in context. Each step is graded, so showing the condition check earns points even if your arithmetic slips later.

Multiple-choice questions hit the conceptual side hard. Expect questions on what "95% confident" actually means, how interval width responds to changes in n or confidence level, what a p-value is the probability of, and identifying Type I and Type II errors in a described scenario. Error and power questions almost always come wrapped in context (a drug trial, a quality inspection), and you have to say what the error means for that situation, not just recite the definition. Interpretation language matters enormously here; a correct calculation with a sloppy interpretation loses points, so practice the exact sentence templates for intervals, confidence levels, p-values, and conclusions.

Essential questions

• How can one sample tell us anything trustworthy about an entire population?
• Why do we report a range of values instead of a single best guess, and what does "confident" really mean?
• How small does a p-value need to be before we stop believing the null hypothesis, and who decides?
• When a test leads us to the wrong conclusion, which kind of mistake is worse, and how can we control it?

Key terms to know

• Population proportion (p): The true fraction of a population with some characteristic, the parameter you're trying to learn about.
• Sample proportion (p̂): The fraction of your sample with the characteristic, your point estimate of p.
• Confidence interval: A range of plausible values for a parameter, built as point estimate ± margin of error.
• Margin of error: How far the sample statistic is likely to vary from the parameter, equal to the critical value times the standard error.
• *Critical value (z)**: The number of standard errors that corresponds to your confidence level (1.96 for 95% confidence).
• Null hypothesis (H₀): The default claim assumed true during the test, always containing equality.
• Alternative hypothesis (Hₐ): The claim you're collecting evidence for, one-sided (< or >) or two-sided (≠).
• p-value: The probability, assuming H₀ is true, of getting a sample result as extreme or more extreme than the one observed.
• Significance level (α): The cutoff for rejecting H₀, set in advance; it equals the probability of a Type I error.
• Type I error: Rejecting a null hypothesis that is actually true, a false positive.
• Type II error: Failing to reject a null hypothesis that is actually false, a false negative.
• Power: The probability that a test correctly rejects a false null hypothesis, equal to 1 minus P(Type II error).
• Large Counts condition: The check that expected successes and failures are each at least 10, justifying a normal model for p̂.
• Pooled proportion (p̂c): The combined success rate from both samples, used in the two-sample test because H₀ assumes the proportions are equal.

Common mix-ups

• Tests use p₀ in the standard error and conditions; intervals use p̂. The test assumes the null is true, so everything in the test is built from the null value. Mixing these up is one of the most common point losses on FRQs.
• "We are 95% confident the interval captures p" is correct. "There is a 95% probability p is in this interval" is wrong, because p is a fixed number; it's either in your interval or it isn't. The 95% describes how often the method works across many samples.
• Pooling only happens in the two-sample test, never in the two-sample interval. The test pools because H₀ claims p₁ = p₂, so there's one shared proportion to estimate. The interval makes no such assumption.
• "Fail to reject H₀" does not mean "H₀ is true." It means the evidence wasn't strong enough. Writing "accept the null" or "this proves" will cost you on the free response.