This free AP Statistics Unit 3 practice test covers Inference for Categorical Data: Proportions in the course revised for 2026–27: estimators, the sampling distributions of a sample proportion and of a difference between two sample proportions, confidence intervals and significance tests for one and two proportions, p-values, Type I and Type II errors and power, and chi-square tests for homogeneity and independence. It has 20 multiple-choice questions with four answer choices, as on the revised exam, and one 10-point free-response question: 30 points in 40 minutes with a calculator. Every explanation shows the working and says why each wrong answer is wrong.
- Question 11 mark · Estimators and biasUsing software, a statistics class drew 1,000 random samples of 60 households each from a large town in which the proportion of households that own a cat is . From each sample the class calculated three different statistics, I, II and III, each intended as an estimator of . The table shows the mean and the standard deviation of the 1,000 values of each statistic. Which statement is best supported by the simulation?
- A.Statistic II appears to be the best estimator of , because its values vary the least from sample to sample, so its estimates tend to be closest to
- B.All three statistics appear to be unbiased estimators of , because each one was calculated from a random sample of the town's households
- C.Statistic III appears to be the best estimator of , because its larger spread makes its values more likely to reach the true value of
- D.Statistics I and III both appear to be unbiased, and Statistic I is the better estimator of because its values vary less from sample to sample
- A.
- Question 21 mark · Sampling distribution of a proportionIn a large region, 20 percent of all households have rooftop solar panels. A random sample of 100 households from the region is selected, and is the proportion of the sampled households that have rooftop solar panels. Using a Normal approximation to the sampling distribution of , which of the following is closest to the probability that is greater than 0.26?
- A.
- B.
- C.
- D.
- A.
- Question 31 mark · Sampling distribution of a proportionAt a large orchard, 6 percent of the apples have a blemish that makes them unsuitable for sale. An inspector selects a random sample of 50 apples from a harvest of many thousands of apples, and is the proportion of blemished apples in the sample. Which of the following best describes the sampling distribution of ?
- A.Skewed to the right with mean 0.06 and standard deviation 0.034, because the expected number of blemished apples is less than 10
- B.Approximately Normal with mean 0.06 and standard deviation 0.0011, because the 50 apples are less than 10 percent of the whole harvest
- C.Approximately Normal with mean 0.06 and standard deviation 0.0011, because the sample size of 50 apples is at least 30
- D.Skewed to the right with mean 3 and standard deviation 1.679, because the expected number of blemished apples is less than 10
- A.
- Question 41 mark · Confidence intervals for a proportionAn internet provider surveyed a random sample of 400 of its many thousands of customers, and 248 of them reported at least one service outage in the past month. Which of the following is a 95 percent confidence interval for the proportion of all the provider's customers who had at least one outage in the past month?
- A.
- B.
- C.
- D.
- A.
- Question 51 mark · Confidence intervals for a proportionA state wildlife agency plans to survey a random sample of licensed anglers to estimate the proportion of all licensed anglers in the state who released at least one fish last season. The agency wants a 95 percent confidence interval with a margin of error of no more than 0.04, and it has no earlier estimate of the proportion. What is the smallest sample size that guarantees this margin of error?
- A.
- B.
- C.
- D.
- A.
- Question 61 mark · Confidence interval interpretationA fitness-club chain asked a random sample of 500 of its members whether they would attend a new 6 a.m. class, and 224 said yes. From these data, a 90 percent confidence interval for the proportion of all the chain's members who would attend the class is . Which of the following is a correct interpretation of the 90 percent confidence level?
- A.In many random samples of 500 members, about 90 percent of the intervals built this way would capture the proportion of all members who would attend
- B.There is a 0.90 probability that the proportion of all the chain's members who would attend the new 6 a.m. class lies between 0.411 and 0.485
- C.In many random samples of 500 members, about 90 percent of the sample proportions would lie between 0.411 and 0.485, the limits of this one interval
- D.We are 90 percent confident that the proportion of the 500 sampled members who would attend the new 6 a.m. class lies between 0.411 and 0.485
- A.
- Question 71 mark · Confidence intervals for a proportionA city parks department surveyed a random sample of 250 residents, and 80 of them had visited a city park in the past week. The 95 percent confidence interval for the proportion of all residents who had visited a park in the past week had a margin of error of about 0.058. The department will take a new random sample and wants a 95 percent confidence interval with a margin of error of about 0.029. If the sample proportion turns out about the same, approximately how many residents should be in the new sample?
- A.
- B.
- C.
- D.
- A.
- Question 81 mark · One-proportion z-testA national report states that 35 percent of adults in a country did volunteer work in the past year. A city official believes that the proportion is lower among adults in her city. She selects a random sample of 220 adults in the city and finds that 66 of them, a proportion of 0.30, did volunteer work in the past year. Which of the following are appropriate hypotheses for testing the official's belief?
- A.and , where is the proportion of the 220 sampled adults who did volunteer work in the past year
- B.and , where is the proportion of all adults in the city who did volunteer work in the past year
- C.and , where is the proportion of all adults in the city who did volunteer work in the past year
- D.and , where is the proportion of all adults in the city who did volunteer work in the past year
- A.
- Question 91 mark · Interpreting p-valuesA meal-kit company believes that more than 25 percent of its subscribers skip at least one weekly delivery each month. In a random sample of 300 subscribers, 89 had skipped at least one delivery in the past month. A test of against , where is the proportion of all the company's subscribers who skipped at least one delivery in the past month, gives a p-value of 0.031. Which of the following is a correct interpretation of this p-value?
- A.The probability that the proportion of all subscribers who skipped at least one delivery in the past month is equal to 0.25 is 0.031
- B.If exactly 25 percent of all subscribers skipped a delivery, the probability of a sample proportion of or more in a random sample of 300 is 0.031
- C.If more than 25 percent of all subscribers skipped a delivery, the probability of a sample proportion of or more in a random sample of 300 is 0.031
- D.If the company rejects on the basis of this sample, the probability that the company is making a Type I error by doing so is 0.031
- A.
- Question 101 mark · One-proportion z-testA ferry company states that 90 percent of its crossings depart on time. A consumer group suspects that the true proportion is lower. In a random sample of 200 of the company's several thousand crossings last year, 172 departed on time. The group tests against at the level, where is the proportion of all the company's crossings last year that departed on time. Which of the following gives the test statistic, the p-value and the correct decision?
- A.z \approx -1.89$ and p-value $\approx 0.030$; reject $H_0LaTeX syntax issue: KaTeX parse error: Can't use function '$' in math mode at position 16: z \approx -1.89$̲ and p-value $\…
- B.z \approx -1.89$ and p-value $\approx 0.059$; fail to reject $H_0LaTeX syntax issue: KaTeX parse error: Can't use function '$' in math mode at position 16: z \approx -1.89$̲ and p-value $\…
- C.z \approx -1.89$ and p-value $\approx 0.970$; fail to reject $H_0LaTeX syntax issue: KaTeX parse error: Can't use function '$' in math mode at position 16: z \approx -1.89$̲ and p-value $\…
- D.z \approx -1.63$ and p-value $\approx 0.052$; fail to reject $H_0LaTeX syntax issue: KaTeX parse error: Can't use function '$' in math mode at position 16: z \approx -1.63$̲ and p-value $\…
- A.
- Question 111 mark · Type I and II errors & powerA water utility monitors a reservoir. Let be the proportion of all water samples from the reservoir in which a contaminant exceeds the safety limit. Each month the utility tests against , and if it rejects it closes the reservoir for treatment. Which of the following describes a Type II error and its consequence?
- A.Finding convincing evidence that when in fact , so the reservoir is closed for a treatment that it does not really need
- B.Not finding convincing evidence that when in fact , so the reservoir stays open and its water is not given treatment
- C.Finding convincing evidence that when in fact , so the reservoir is closed and its water is given the treatment
- D.Not finding convincing evidence that when in fact , so the reservoir stays open while its water needs treatment
- A.
- Question 121 mark · Type I and II errors & powerA town council member will test whether more than half of the town's adults favor a ban on gas-powered leaf blowers. She plans to survey a random sample of 120 adults and test against at the level. Suppose that, in fact, 56 percent of the town's adults favor the ban. Which of the following changes to her plan would increase the power of the test, with everything else unchanged?
- A.Surveying a random sample of 300 of the town's adults instead of a random sample of 120 adults
- B.Carrying out the test at the significance level instead of at the level
- C.Using the two-sided alternative instead of the one-sided alternative
- D.Testing against instead of testing against the null value 0.50
- A.
- Question 131 mark · Sampling distribution of a difference in proportionsAt a large university with thousands of students in each of its schools, 40 percent of engineering students and 25 percent of business students own a tablet computer. Independent random samples of 100 engineering students and 150 business students are selected, and is the difference between the proportions of tablet owners in the two samples. What are the mean and the standard deviation of the sampling distribution of ? (Standard deviations are rounded to three decimal places.)
- A.Mean 0, standard deviation 0.084
- B.Mean 0.15, standard deviation 0.034
- C.Mean 0.15, standard deviation 0.060
- D.Mean 0.15, standard deviation 0.084
- A.
- Question 141 mark · Confidence intervals for a difference in proportionsIndependent random samples of first-year students were taken at two large universities, each with more than 4,000 first-year students. At University N, 204 of 300 sampled students lived on campus; at University S, 145 of 250 sampled students lived on campus. Which of the following is a 90 percent confidence interval for , the difference between the proportions of all first-year students at the two universities who live on campus?
- A.
- B.
- C.
- D.
- A.
- Question 151 mark · Confidence intervals for a difference in proportionsA sports physiotherapist randomly assigned 360 runners with a sprained ankle to a new balance-training program or to standard care, 180 runners to each. After eight weeks, 126 runners in the balance-training group and 117 in the standard-care group had fully recovered. A 95 percent confidence interval for , the difference between the proportions of such runners who would fully recover within eight weeks with each treatment, is . Which conclusion is best supported?
- A.The data give convincing evidence that the balance-training program gives a higher recovery rate, because most of the interval is positive
- B.The data give convincing evidence that the two treatments give the same recovery rate, because the interval contains the value 0
- C.The data give convincing evidence that the balance-training program gives a higher recovery rate, because the observed difference is above 0
- D.The data do not give convincing evidence that the recovery rates differ, because 0 is one of the plausible values in the interval
- A.
- Question 161 mark · Two-proportion z-testA researcher selected independent random samples of households in two large towns. In Town A, 45 of 120 sampled households own a dog; in Town B, 54 of 180 sampled households own a dog. She plans a test of against . Which of the following correctly gives the combined (pooled) sample proportion and the check of the normality condition?
- A.; the condition is not met, since , which is less than 10
- B.; the condition is met, since , , and are all at least 10
- C.; the condition is met, since each of the two samples contains more than 30 households and both towns are large
- D.; the condition is met, since , , and are all at least 10
- A.
- Question 171 mark · Two-proportion z-testIndependent random samples of 500 households were taken in each of two large cities. In City M, 60 of the sampled households had a backyard vegetable garden; in City N, 42 did. A researcher tests against at the level, where and are the proportions of all households in each city with a backyard vegetable garden. Which of the following gives the test statistic, the p-value and a correct conclusion?
- A.and p-value ; there is convincing evidence that the proportions of households with a garden differ in the two cities
- B.and p-value ; there is not convincing evidence that the proportions of households with a garden differ in the two cities
- C.and p-value ; there is convincing evidence that the proportions of households with a garden are equal in the two cities
- D.and p-value ; there is convincing evidence that the proportions of households with a garden differ in the two cities
- A.
- Question 181 mark · Chi-square tests for homogeneity and independenceA coffee chain wants to know whether customers' usual drink size differs among its three locations. It selects an independent random sample of 120 customers at each location and records each customer's usual drink size: small, medium or large. Which of the following gives the appropriate test and null hypothesis?
- A.A chi-square test for independence, with : location and usual drink size are independent in a single population of the chain's customers
- B.A two-sample z-test for proportions, with : the proportion of customers who usually order a large drink is the same at two of the locations
- C.A chi-square test for homogeneity, with : the distribution of usual drink size is the same for the customers of all three locations
- D.A chi-square test for homogeneity, with : the distributions of usual drink size are different for the customers of the three locations
- A.
- Question 191 mark · Chi-square tests for homogeneity and independenceA random sample of 300 homes in a large city was classified by type of home and main heating fuel, as shown in the table. A chi-square test will be carried out to investigate whether type of home and main heating fuel are associated for homes in the city. What are the expected count for apartments heated mainly by gas and the degrees of freedom of the test?
- A.Expected count 50; 5 degrees of freedom
- B.Expected count 56; 2 degrees of freedom
- C.Expected count 56; 5 degrees of freedom
- D.Expected count 64; 2 degrees of freedom
- A.
- Question 201 mark · Chi-square tests for homogeneity and independenceA researcher randomly assigned 200 volunteers with seasonal allergies to use nasal spray A or nasal spray B for two weeks, 100 volunteers to each spray, and recorded each volunteer's improvement. The results are shown in the table. A chi-square test is carried out at the level to investigate whether the distribution of improvement differs for the two sprays. Which of the following gives the chi-square statistic, the p-value and a correct conclusion?
- A.and p-value ; there is convincing evidence that the distribution of improvement is the same for the two sprays
- B.and p-value ; there is convincing evidence that the distribution of improvement differs for the two sprays
- C.and p-value ; there is convincing evidence that the distribution of improvement is the same for the two sprays
- D.and p-value ; there is not convincing evidence that the distribution of improvement differs for the two sprays
- A.
- Question 2110 marks · Two-proportion test, conditions & errorsComparing two reminder methods. A county health department wanted to know whether a text-message reminder leads to a higher flu-vaccination rate among older adults than a postcard reminder sent by mail. From the county's list of 15,000 adults aged 65 and over who had not yet had a flu shot that season, the department selected a random sample of 600 adults. It then randomly assigned 300 of them to receive a text-message reminder and the other 300 to receive a postcard reminder. The table shows how many adults in each group got a flu shot within four weeks. Use the 0.05 significance level, and give p-values to three decimal places.(a)[2]Define the parameters of interest, and state the null and alternative hypotheses for investigating whether a text-message reminder leads to a higher flu-vaccination rate than a postcard reminder.(b)[2]Identify the appropriate inference procedure, and verify that the conditions for it are met.(c)[2]Calculate the test statistic and the p-value.(d)[2]State a conclusion at the 0.05 significance level, in the context of the study.(e)[2]Based on your conclusion in part (d), which type of error, Type I or Type II, could the department have made? Describe that error in the context of the study, and give one possible consequence of it for the department.
Finished?
Focus on specific topics (optional)
AP Statistics exam tips
- Answer every inference question in four labelled steps. State: define the parameter or parameters in words, in context, and write the hypotheses. Plan: name the procedure in full (for example, "two-sample z-interval for a difference between population proportions") and check its conditions. Do: write the formula with the numbers substituted, then the result. Conclude: compare the p-value with the significance level, make a decision about the null hypothesis and say in context what it means. Calculator syntax on its own, such as 2-PropZTest with no named procedure or numbers, risks losing credit.
- Check conditions with numbers from the question, not with labels. Say how the data were randomized (a random sample, independent random samples or random assignment); when sampling without replacement, show the 10 percent comparison, such as 150 ≤ 0.10 × 4,000; and show the normality products themselves, such as 150(0.24) = 36 ≥ 10. The course checks the observed successes and failures for a confidence interval, uses the null-hypothesis value for a one-sample test and the combined proportion for a two-sample test, and requires every expected count of a chi-square test to be at least 5. "SRS" or "large counts" with no numbers is not a check.
- Say which distribution you mean. The population, the values in one sample and the sampling distribution of a statistic over many samples are three different things, and a sentence such as "it gets less variable as n increases" leaves the reader unsure which one "it" is. When you describe a sampling distribution, give its shape with a reason, its center and its spread, in context. Write every interpretation of an interval, a confidence level or a p-value in the words of the question.
- Keep calculations traceable. Write the standard error with the numbers substituted before you use it, keep unrounded values in the calculator until the last step, and round probabilities and p-values to three decimal places. For a chi-square test, report the expected counts, the degrees of freedom, the statistic and the p-value together. If a multiple-choice answer you calculate is not among the options, recheck the critical value, the tail area and the square root before you guess.
More free AP practice exams
- AP Microeconomics Practice Exam
- AP Macroeconomics Practice Test
- AP Macroeconomics Free-Response (FRQ) Practice Exam
- AP Macroeconomics Multiple Choice Questions with Answers
- AP Microeconomics Free-Response Practice Questions (FRQ) with Answers
- AP Macro Unit 2 Practice Test: Economic Indicators and the Business Cycle
Frequently asked questions
Is this AP Stats Unit 3 practice test really free?
Yes. The full exam and the print-ready PDF are free with no signup. The worked answer key is free too: enter your email on the page and we email you the answer-key PDF, with the point-by-point scoring of the free-response question, and unlock the solutions on screen. ExamTeX makes money from its exam generator, which turns your own class notes into practice exams like this one.
What does Unit 3 cover on the revised AP Statistics exam?
College Board revised AP Statistics for the 2026–27 school year into five units. Unit 3, Inference for Categorical Data: Proportions, is weighted at 15–25 percent of the multiple-choice section. Its 15 topics run from estimators and the sampling distributions of a sample proportion and of a difference between two sample proportions, through confidence intervals and significance tests for one proportion and for a difference between two proportions, p-values, Type I and Type II errors and power, to chi-square tests for homogeneity and independence. This test covers all of them. The Normal distribution and the central limit theorem are taught in Unit 2 and inference for means in Unit 4, so they are not tested here. The revised exam has 42 multiple-choice questions with four answer choices each and 4 free-response questions worth 10 points each, one of which is an inference question, which is why this test uses four-choice questions and a 10-point free-response question.
My class or textbook calls this material Unit 6 or Unit 8. Is it the same?
Mostly, yes. In the nine-unit numbering used before the 2026–27 revision, this material was spread over three units: Unit 6, Inference for Categorical Data: Proportions (intervals and tests for one and two proportions, errors and power); Unit 8, Inference for Categorical Data: Chi-Square (the tests for homogeneity and independence; that unit's goodness-of-fit test is not in the revised course framework, and this test does not cover it); and the parts of Unit 5, Sampling Distributions, on estimators and on the sampling distributions of a sample proportion and of a difference in sample proportions. The rest of old Unit 5 moved: the Normal distribution and the central limit theorem to the revised Unit 2, and the sampling distributions of sample means to Unit 4.
Can I generate more AP Stats inference questions like these?
Yes. Upload your own notes, slides or a past paper to ExamTeX and it generates a print-ready practice exam in the same format in about a minute, with an answer key included. The free tier includes 1 exam a month.
Can teachers use this exam in class?
Yes. The PDF is print-ready (A4, standard exam layout) and you are welcome to photocopy it for classroom use. Attribution is appreciated but not required.
ExamTeX