Worksheets · AP Statistics · Unit 3

AP Statistics Unit 3 FRQ Practice: Inference for Proportions and Simulation

Unit 3: Inference for Categorical Data: Proportions

Two exam-style free-response questions on inference for proportions, each scored out of 10 points with a scoring guide and a printable PDF with room to write, free to use in class. Good practice before the Unit 3 Progress Check in AP Classroom. Multiple choice for this unit: Unit 3 worksheet.

Sigma Prep · sigmaprep.io/worksheets/ap-statistics

Name

AP Statistics Unit 3 Free Response: Inference for Proportions and Simulation

Date

Directions: Show all your work. Indicate clearly the methods you use, because you will be scored on the correctness of your methods as well as on the accuracy and completeness of your results and explanations. Answer in the context of the question. Each question is worth 10 points, and each point is earned on its own. A graphing calculator with statistical capabilities may be used, and the exam’s formula sheet and tables apply.

Questions 1 and 2 practice the kinds of questions in Questions 3 and 4 of the exam’s free-response section. The exam’s Question 1 (data collection) carries little of Unit 3’s official free-response points, so the set has no question in its style, and few of its official free-response points are in Question 2.

Free response

Questions 1–2
  1. 1.
    A hospital network compared how often nurses on day shifts and on night shifts report feeling rested. The network employs about 6,000 day-shift nurses and about 3,500 night-shift nurses. Independent random samples of 400 day-shift nurses and 100 night-shift nurses were selected. Of the 400 day-shift nurses, 176 report feeling rested; of the 100 night-shift nurses, 39 report feeling rested. The network's wellness team wants to know whether the proportion of all day-shift nurses in the network who report feeling rested is greater than the proportion of all night-shift nurses in the network who report feeling rested.
    1. (A)
      Identify the explanatory variable and the response variable in the study, and classify each as categorical or quantitative.
    2. (B)
      Identify the appropriate inference procedure for investigating the question.
    1. (C)
      (i) State the null and alternative hypotheses, and define the parameters. (ii) Verify the conditions for the test. (iii) Calculate the test statistic and the P-value.
    2. (D)
      Based on the P-value, what conclusion should be made at the α=0.01\alpha = 0.01 level? Justify your answer.
  2. 2.
    A factory cut bolts whose lengths, when the cutter worked properly, were approximately normal with mean 80 millimeters and standard deviation 0.4 millimeters. An engineer suspected that the lengths had become more variable and measured a random sample of 8 bolts. The lengths, in millimeters, were 79.8, 79.8, 80.8, 80.6, 80.3, 79.4, 79.7, 80.0. To judge whether the sample range is unusually large, the engineer simulated 100 random samples of 8 bolts from a normal population with mean 80 and standard deviation 0.4 millimeters, and recorded the range of each sample. The dotplot below shows the 100 simulated ranges.
    {"groups":[{"name":"Simulated ranges","values":[0.4,0.9,1.3,1,1.5,0.9,0.6,0.8,1.4,1.2,0.9,0.6,0.9,0.9,0.6,1,1.3,2.1,1.3,1.2,1.3,1.3,1,1.5,0.8,1.2,1.3,1.8,1.2,1.1,0.8,1.1,1.6,0.9,1.3,1.5,0.9,1.5,1.3,1.3,1.1,1.3,0.9,0.9,1,1.1,1,0.4,1.2,1.3,1.8,2,1.4,1.5,0.7,1.5,1.5,0.7,0.8,1.3,1.1,0.8,0.3,0.9,1.2,1.1,1.2,1.1,1.2,1,1.2,1.4,0.7,0.9,1,1.1,1.6,1.5,0.8,0.5,1.3,1.4,0.8,0.7,0.5,0.8,1.3,1.9,1.5,1.4,0.7,0.8,1.3,1.1,0.6,1.8,0.7,0.9,1,1.2]}],"unit":0.1,"axis":{"lo":0,"hi":2.5,"step":0.5,"dp":1}}0.00.51.01.52.02.5Sample range (millimeters)
    1. (A)
      State the null and alternative hypotheses that the engineer is testing, in terms of the standard deviation σ\sigma of the lengths.
    2. (B)
      Describe how one dot in the dotplot was obtained.
    1. (C)
      (i) Calculate the range of the sample. (ii) Use the dotplot to estimate the P-value of the test. Count a simulated range equal to the observed range as at least as large.
    2. (D)
      What should the engineer conclude at the α=0.05\alpha = 0.05 level? Explain.
    1. (E)
      Based on the simulation, what is the smallest sample range that would lead the engineer to conclude, at the α=0.05\alpha = 0.05 level, that the lengths have become more variable? Explain.
    2. (F)
      A second simulation took 100 random samples of 8 bolts from a normal population with standard deviation 0.7 millimeters. Of the 100 sample ranges, 54 were at least as large as the rejection cutoff from part (E). (i) Estimate the power of the test against σ=0.7\sigma = 0.7. (ii) Interpret this power in context.
    1. (G)
      How would the power against σ=0.7\sigma = 0.7 change if the engineer used random samples of 16 bolts instead of 8? Explain.
    2. (H)
      Explain why the engineer used a simulation instead of a zz or tt procedure to test this claim.

      End of the free-response questions

Free AP Statistics worksheets at sigmaprep.io/worksheets/ap-statisticsOriginal questions written by Sigma Prep. AP is a registered trademark of the College Board, which is not affiliated with this sheet.

Scoring guide

Each point is earned on its own: a response earns a point when it meets that point’s requirement, whatever it earned on the others. The model answers are one complete response; “Also acceptable” names other responses that earn the same point.

Question 1: Comparing two proportionsLike exam Question 3 (Inference) · 10 points

(A) 2 points

  1. Point 1. Identifies the explanatory variable (the group or treatment) and classifies it as categorical. (Practice 2)

    Model answer: The explanatory variable is shift (day or night); it is categorical.

  2. Point 2. Identifies the response variable and classifies it as categorical. (Practice 2)

    Model answer: The response variable is whether the nurse reports feeling rested (yes or no); it is categorical.

(B) 1 point

  1. Point 3. Names a two-sample z-test for a difference between two population proportions. (Practice 2)

    Model answer: A two-sample zz-test for a difference between two population proportions.

    Also acceptable: “2-PropZTest”, with the parameters made clear.

(C) 5 points

  1. Point 4. Correct hypotheses about p1−p2p_1 - p_2. (Practice 2)

    Model answer: (i) H0:p1−p2=0H_0: p_1 - p_2 = 0 and Ha:p1−p2>0H_a: p_1 - p_2 > 0.

    Also acceptable: H0:p1=p2H_0: p_1 = p_2 with the matching HaH_a.

    Also acceptable: The other order of subtraction (p2−p1p_2 - p_1, with the direction of HaH_a reversed), used consistently with the parameters.

  2. Point 5. Defines both parameters in context, as population (or true treatment) proportions. (Practice 2)

    Model answer: Here p1p_1 is the proportion of all day-shift nurses in the network who report feeling rested and p2p_2 is the proportion of all night-shift nurses in the network who report feeling rested.

  3. Point 6. Independent random samples stated, and the 10% condition checked for each population with its size. (Practice 2)

    Model answer: (ii) Random: independent random samples of day-shift nurses and night-shift nurses. For the 10% condition, 400 is less than 10% of the 6,000 day-shift nurses, and 100 is less than 10% of the 3,500 night-shift nurses.

  4. Point 7. Large counts checked with the pooled proportion (all four counts at least 10). (Practice 2)

    Model answer: Large counts: with p^c=176+39400+100=0.4300\hat{p}_c = \frac{176 + 39}{400 + 100} = 0.4300, the expected counts 172.00, 228.00, 43.00, 57.00 are all at least 10.

    Also acceptable: The observed counts 176, 224, 39 and 61, all at least 10.

  5. Point 8. Correct test statistic and P-value. (Practice 3)

    Model answer: (iii) p^1=0.4400\hat{p}_1 = 0.4400, p^2=0.3900\hat{p}_2 = 0.3900, z=0.90z = 0.90 and P-value =0.1832= 0.1832.

    Also acceptable: The values from technology (2-PropZTest), to rounding.

    Also acceptable: The table route: z=0.90z = 0.90 (rounded to two places) gives a P-value of 0.1841 from the standard normal table.

(D) 2 points

  1. Point 9. Compares the P-value with α\alpha and makes the correct decision. (Practice 4)

    Model answer: Because the P-value of 0.1832 is greater than α=0.01\alpha = 0.01, do not reject H0H_0.

    Also acceptable: Any other correct, complete justification in context earns the point.

    Also acceptable: Carried-forward credit: a response that uses its own value from (C)(iii) (here 0.1832) earns this point when its answer here is correct for that value.

  2. Point 10. Conclusion in context about the population proportions, in terms of HaH_a. (Practice 4)

    Model answer: There is not convincing statistical evidence that the proportion of all day-shift nurses in the network who report feeling rested is greater than the proportion of all night-shift nurses in the network who report feeling rested.

    Also acceptable: Carried-forward credit: a response that uses its own result from Point 9 earns this point when its answer here is correct for that result.

Question 2: A simulation, a P-value and powerLike exam Question 4 (Practices 2, 3 and 4) · 10 points

(A) 1 point

  1. Point 1. Correct hypotheses about the population SD: the standard deviation of a properly working process (H0H_0) against greater variability (HaH_a). (Practice 2)

    Model answer: H0:σ=0.4H_0: \sigma = 0.4 millimeters (the process works properly) and Ha:σ>0.4H_a: \sigma > 0.4 millimeters, where σ\sigma is the standard deviation of the lengths of all bolts it produces now.

(B) 1 point

  1. Point 2. A random sample of the same size from the population that H0H_0 describes (normal, the stated mean and SD), and its range (maximum minus minimum) recorded. (Practice 2)

    Model answer: Select a random sample of 8 values from a normal distribution with mean 80 and standard deviation 0.4 millimeters, and record its range, the largest value minus the smallest.

    Also acceptable: Any other correct, complete description in context earns the point.

(C) 2 points

  1. Point 3. Correct observed range. (Practice 3)

    Model answer: (i) Compute 80.8−79.4=1.480.8 - 79.4 = 1.4 millimeters.

  2. Point 4. The proportion of simulated ranges at least as large as the observed range. (Practice 3)

    Model answer: (ii) Of the 100 simulated ranges, 22 are at least 1.4, so the estimated P-value is 22100=0.22\frac{22}{100} = 0.22.

    Also acceptable: Carried-forward credit: a response that uses its own value from (C)(i) (here 1.4) earns this point when its answer here is correct for that value.

(D) 1 point

  1. Point 5. The decision from the estimated P-value, with a conclusion about the variability in context. (Practice 4)

    Model answer: The estimated P-value of 0.22 is greater than 0.05, so there is not convincing evidence that the standard deviation of the lengths is greater than 0.4 millimeters.

    Also acceptable: Any other correct, complete justification in context earns the point.

    Also acceptable: Carried-forward credit: a response that uses its own value from (C)(ii) (here 0.22) earns this point when its answer here is correct for that value.

(E) 1 point

  1. Point 6. The smallest simulated value with at most 5% of the dots at or above it. (Practice 3)

    Model answer: A cutoff of 1.9 millimeters: only 3 of the 100 simulated ranges are 1.9 or more (at most 5%), while 6 are at least the next smaller value.

    Also acceptable: Any equivalent statement of the same rule, such as “any range greater than 1.8 millimeters” (the next smaller simulated value), which gives the same 3 of 100.

    Also acceptable: Any other correct, complete justification in context earns the point.

(F) 2 points

  1. Point 7. Power estimated as the proportion of samples that would lead to rejecting H0H_0. (Practice 3)

    Model answer: (i) Compute 54100=0.54\frac{54}{100} = 0.54.

  2. Point 8. Power as the probability of correctly concluding that the variability increased when the true SD is the alternative value. (Practice 4)

    Model answer: (ii) If the standard deviation of the lengths is actually 0.7 millimeters, there is about a 0.54 probability that a sample of 8 bolts will lead the engineer to correctly conclude that the variability has increased.

    Also acceptable: Any other correct, complete interpretation in context earns the point.

    Also acceptable: Carried-forward credit: a response that uses its own value from (F)(i) (here 0.54) earns this point when its answer here is correct for that value.

(G) 1 point

  1. Point 9. Power increases with the sample size, because the sampling distribution of the statistic under each SD is less variable, so the two overlap less. (Practice 4)

    Model answer: It would increase: with larger samples, the range varies less from sample to sample, so the ranges from a population with standard deviation 0.7 millimeters overlap less with those from 0.4 millimeters. The cutoff for samples of the larger size would be found again from a new simulation that uses the standard deviation in the null hypothesis, and a larger share of samples from the alternative would land beyond that new cutoff.

    Also acceptable: Any other correct, complete justification in context earns the point.

(H) 1 point

  1. Point 10. The claim is about variability (the sample range), whose sampling distribution has no simple formula or normal shape, so it is simulated; the zz and tt procedures are for proportions and means. (Practice 2)

    Model answer: The zz procedures are for proportions and the tt procedures are for means, and both rely on a sampling distribution with a known, roughly normal shape. The claim is about variability, measured here by the sample range, whose sampling distribution has no simple formula and is skewed, so simulation is used to find it.

    Also acceptable: Saying only that the sample range has no known (or normal) sampling distribution, so its distribution has to be simulated.

    Also acceptable: Any other correct, complete justification in context earns the point.

What does the AP Statistics free response ask on Unit 3?

Unit 3 carries about a fifth of the AP Statistics free response: 195 of the 1,040 points on the 124 official free-response questions we counted (18.7%), and 8 of the 40 points in the four 2026 sample questions.

  • Question 3 is a full inference question. On the 2027 exam, Question 3 asks for a significance test or a confidence interval, from the conditions to the conclusion. Of the 17 official free-response questions from 2012 to 2026 that fit Question 3, 12 had a Unit 3 part, 5 of them since 2019; the 2026 sample Question 3 is a two-proportion test.
  • Proportions are the usual setting. Inference for proportions was the setting of 12 official free-response questions from 2012 to 2026, 5 of them since 2019 and most recently in 2025, and a chi-square test of 3, most recently in 2017.
  • Question 4 can build on Unit 3. 10 official free-response questions from 2012 to 2026 that fit the Question 4 slot had a Unit 3 part, such as a Type I or Type II error, power or a P-value.
  • Every point is earned on its own. From 2027 each free-response question is worth 10 points, scored analytically: each point has its own requirement, and a later point does not depend on an earlier one.

Unit 3 has no data-collection question, and nearly all of its free-response points are in Questions 3 and 4, so this set is practice for those two questions. Every question and scoring guide is original.

Counted from the official AP Statistics free-response questions: the 2012–2026 exams, the practice exams and the 2026 course-description samples. No official question is reproduced here.

How often does the AP Statistics free response use each Unit 3 setting?

Settings of the questions on this set, and how many official AP Statistics free-response questions used each, 2012–2026
SettingQuestions on this setOfficial FRQs, 2012–2026Of them since 2019Last on the exam
Comparing two proportionsInference for proportions · exam Question 31 of 21252025
A simulation, a P-value and powerQuestion 4 with a Unit 3 part · exam Question 41 of 21042025

Counts are of the 84 operational free-response questions, 2012–2026 (no 2020 exam was released), by the setting of the whole question. The Question 4 row counts questions that fit the exam’s new Question 4 and have a Unit 3 part.

Questions about this free-response set

Unit 3, Inference for Categorical Data: Proportions, has 15 topics in the College Board's 2026 course and exam description: estimators and bias; the sampling distribution of a sample proportion; constructing and interpreting a confidence interval for a population proportion and using it to judge a claim; setting up and carrying out a significance test for a proportion, with P-values (including P-values from a simulation); Type I and Type II errors and power; the same for the difference of two proportions (the sampling distribution of the difference, the two-proportion z-interval and z-test); and chi-square tests for homogeneity and independence in a two-way table. The course description weights the unit at 15–25% of the multiple-choice section.

Each of the four free-response questions is worth 10 points, scored analytically: every point has its own requirement (a variable named and classified, a sampling method identified, a comparison made in context with values), and each is earned on its own. The scoring guide here lists all 10 points of every question, with a model answer for each and the other responses that also earn it. The free-response section is half of the exam score.

The 2027 exam's Section II has four questions: Question 1 on Practices 1 and 2 (the investigative question and collecting data), Question 2 on Practices 3 and 4 (analyzing data and drawing conclusions), Question 3 on inference, and Question 4 on Practices 2, 3 and 4. The exam's Question 1 (data collection) carries little of Unit 3's official free-response points, and few of its official free-response points are in Question 2, so this set has two questions, in the style of Questions 3 and 4.

Yes. A graphing calculator with statistical capabilities is allowed throughout the exam, and the exam supplies a formula sheet and tables. From 2027 the free-response answers are typed in Bluebook; this set prints with room to write so it can be done on paper in class.

We tagged every official free-response question from 2012 to 2026 (and the four 2026 course-description samples) by its setting and by the exam question it would fill from 2027. Each question here is in a setting the exam uses for that question: a two-proportion z-test, and a P-value and the power of a test estimated from a simulation.

No. Every question here is original, written by Sigma Prep in the style of the exam: the settings, data and wording are our own, and no official question or scoring guideline is copied or reworded. Every number and every claim a scoring point makes (the method, the bias, the direction, the scope of a conclusion) was checked by an independent computer check before the set was published.

AP is a registered trademark of the College Board, which was not involved in making these questions and does not endorse them. All questions and scoring guides are original to Sigma Prep. Frozen 2026-11-20.