Skip to content

Free-response practice

FRQ 1

10 points2 original prompts4 scored examples

FRQ 1 is the first of four free-response questions on the May 2027 AP Statistics exam. The free-response section lasts 90 minutes and counts for 50% of your score. Each of the four questions is worth 10 points, so FRQ 1 alone carries about one-eighth of your whole exam score. Plan on roughly 20 minutes for it, which leaves time for the inference question later in the section.

On this page · 6 sections

This question focuses on the first two course practices: choosing a method or study design, and analyzing data. It most often draws on exploratory data analysis and data collection from Unit 1. You describe and compare distributions, compute a summary value such as a z-score or an outlier fence, and judge or design a way of collecting data. But it can instead be built on a two-way table or on a scatterplot with regression output. With a two-way table, compare conditional relative frequencies (percents within each group), not raw counts. With a scatterplot, describe direction, form, strength and unusual features in context, and interpret the slope, y-intercept and r in context. Expect a short stimulus, usually a table, a graph or computer output, followed by three or four lettered parts.

Each part carries its own points, and the parts add up to 10. Readers score what you write, not what you meant. A correct number with no work shown, or with no link to the story in the stem, often earns nothing. Two habits earn most of the points. First, put every answer in context: say "seedling height in centimeters," not "the data." Second, use comparative language: "Mix B's median height is 2.4 cm greater than Mix A's" earns credit, while two medians listed side by side do not.

You may use a graphing calculator, and you get the formula sheet and tables. But calculator syntax such as "1-Var Stats" is not work. Write the formula with your numbers plugged in, then the answer.

The rubric, in plain terms

10 points, earned line by line. Readers score what is on the page, nothing else.

Identify the method, variable or study type, and justify it2 pts

1 point for a correct identification, such as naming the explanatory and response variables or calling the study observational or experimental. 1 point for a justification tied to this study's features, such as "the researchers assigned the treatments" or "nothing was imposed on the students, so no cause-and-effect claim is allowed."

Compare distributions in context3 pts

1 point for comparing centers, 1 point for comparing variability, and 1 point for comparing shape or naming an unusual feature such as an outlier. Each point needs comparative wording ("greater than," "about the same as"), not a list of values. Each must also be framed in the variable's context; a response with no context anywhere loses one point.

Calculate and interpret a summary measure2 pts

1 point for a correct calculation with the work shown, such as the fence formula or z-score formula with the actual numbers in it. 1 point for a conclusion or interpretation in context that follows from that calculation, for example "2 standard deviations above the mean sleep for students without a first-period class."

Design or critique a data-collection plan3 pts

1 point for a random process described clearly enough to repeat: label the units, name the random tool, and say how many go where. 1 point for building the key structure correctly, such as randomizing separately within each block or sampling separately within each stratum. 1 point for explaining the statistical benefit in context, usually that it reduces variability by grouping units that respond alike.

Practice prompts

Prompt 1

Potting Mixes and Seedling Height

A garden center tested two potting mixes. Forty tomato seedlings of the same variety were randomly assigned, 20 to Mix A and 20 to Mix B. After four weeks, a worker measured each seedling's height in centimeters.

Mix A (n = 20): mean 18.5, SD 2.1, minimum 14.2, Q1 17.0, median 18.6, Q3 20.1, maximum 22.4 Mix B (n = 20): mean 21.4, SD 3.0, minimum 15.8, Q1 19.4, median 21.0, Q3 22.3, maximum 29.0

  1. (a)Identify the explanatory variable and the response variable in this study. Explain why this study is an experiment rather than an observational study. [2 pts]
  2. (b)Compare the distributions of seedling height for Mix A and Mix B. [3 pts]
  3. (c)Use the 1.5 × IQR rule to determine whether the tallest Mix B seedling, at 29.0 cm, is an outlier. Show your work. [2 pts]
  4. (d)The 40 seedlings were kept on two shelves: 20 on a sunny shelf by a window and 20 on a shady shelf. Describe how the garden center could use blocking to assign the two mixes to the seedlings. Explain how blocking would help in this context. [3 pts]
Prompt 2

Sleep and First-Period Classes

A high school has about 1,600 students, roughly 400 in each grade. The student council selected a simple random sample of 120 juniors. Each student reported how many hours they slept the previous school night and whether their schedule includes a first-period class.

First-period class (n = 70): mean 6.4, SD 1.0, minimum 3.5, Q1 5.8, median 6.5, Q3 7.1, maximum 8.6 No first-period class (n = 50): mean 7.4, SD 0.9, minimum 5.2, Q1 6.8, median 7.5, Q3 8.0, maximum 9.6

  1. (a)Is this study an observational study or an experiment? Based on your answer, can the council conclude that having a first-period class causes juniors to sleep less? Explain. [2 pts]
  2. (b)Compare the distributions of hours of sleep for juniors with and without a first-period class. [3 pts]
  3. (c)Juan has no first-period class and slept 9.2 hours. Calculate the z-score for Juan's sleep within his group, and interpret it in context. [2 pts]
  4. (d)The principal wants to estimate the mean nightly sleep of all 1,600 students. Describe how to select a stratified random sample of 200 students using grade level as the strata. Explain one advantage of stratifying by grade instead of taking a simple random sample of 200 students. [3 pts]

Scored examples

Two responses per prompt — one near the top of the rubric, one in the middle — with every point accounted for. The mid response earns its score through the mistakes students actually make.

Potting Mixes and Seedling Height · high-scoring response10/10

(a) Explanatory variable: which potting mix the seedling got (A or B). Response variable: seedling height in cm after 4 weeks. It's an experiment because the garden center imposed the treatment, randomly assigning each seedling to a mix, instead of just observing seedlings that already grew in one mix.

(b) Center: Mix B seedlings were typically taller. The median height for Mix B (21.0 cm) is 2.4 cm greater than for Mix A (18.6 cm). Variability: the middle 50% of heights is about equally spread for both mixes (IQR A = 3.1 cm, IQR B = 2.9 cm). Mix B has a much larger range (13.2 cm vs 8.2 cm), but that is mostly due to one very tall seedling. Shape: Mix A heights look roughly symmetric, since the median is near the middle of the box and the mean (18.5) is close to the median (18.6). Mix B heights are roughly symmetric in the middle, with the median near the center of the box, but they have a high outlier at 29.0 cm, which pulls the mean (21.4) above the median (21.0).

(c) IQR for B = 22.3 − 19.4 = 2.9. Upper fence = Q3 + 1.5(IQR) = 22.3 + 1.5(2.9) = 22.3 + 4.35 = 26.65 cm. Since 29.0 > 26.65, the 29.0 cm seedling is an outlier among Mix B heights.

(d) Block by shelf, since sunlight probably affects growth. On the sunny shelf, number the 20 seedlings 1-20. Use a random number generator to pick 10 different numbers from 1-20, ignoring repeats. Those 10 seedlings get Mix A and the other 10 get Mix B. Then do the same thing separately on the shady shelf. Blocking removes the sunlight-caused variation in height from the comparison of the mixes, so it's easier to tell whether Mix B really grows taller seedlings. It also makes sure each mix gets 10 sunny and 10 shady seedlings, so sunlight can't be mixed up with the effect of the potting mix.

How it scored

2/2a — Both variables are named in context, and the justification points to the imposed, randomly assigned treatment, which is exactly what separates an experiment from an observational study.
3/3b — Center, variability and shape are each compared with comparative wording and in context. Using the IQR for spread, and saying the range is inflated by the tall seedling, shows the student knows which measures resist outliers.
2/2c — The fence formula appears with the numbers substituted, and the conclusion is stated in context. Both points are earned.
3/3d — The random process can be repeated (labels, generator, no repeats, 10 per mix), randomization happens separately within each shelf, and the benefit is tied to sunlight and growth. All three points are earned.
Potting Mixes and Seedling Height · mid-scoring response5/10

(a) The explanatory variable is the potting mix and the response is how tall the seedlings got. This is an experiment because the seedlings were randomly assigned to the mixes.

(b) Mix B has a higher center than Mix A, because its median seedling height is 21.0 cm compared to 18.6 cm for Mix A. Mix B seedling heights are more spread out than Mix A heights because the standard deviation is 3.0 cm vs 2.1 cm. So Mix B is the better mix.

(c) Mean + 2SD = 21.4 + 2(3.0) = 27.4. Since 29.0 is more than 27.4 it is an outlier.

(d) Split the seedlings into a sunny block and a shady block. In each block, randomly choose 10 seedlings to get Mix A and give the rest Mix B. Blocking reduces bias so the results are more accurate.

How it scored

2/2a — Both variables are identified, and random assignment of treatments is correctly given as the reason this is an experiment.
2/3b — Center and variability are both compared in context with comparative wording, so those points are earned. Shape is never mentioned, and neither is the high value at 29.0 cm, so the shape point is lost. "Mix B is the better mix" goes beyond what the question asked and earns nothing.
0/2c — The question asked for the 1.5 × IQR rule. Mean plus two standard deviations is a different rule, and box-plot fences never use the SD. The answer happens to match, but the method is wrong, so neither point is earned.
1/3d — Randomizing within each shelf earns the structure point. "Randomly choose" with no labels or random tool cannot be repeated, so the process point is lost. Blocking reduces variability from a known source like sunlight, not bias, so the benefit point is lost.
Sleep and First-Period Classes · high-scoring response9/10

(a) It's an observational study. The council just recorded which juniors already have a first-period class; nobody was assigned to one. So they cannot conclude that a first-period class causes less sleep. For example, students who take first period might also have more activities or jobs, and that could be what cuts their sleep.

(b) Center: juniors with no first-period class typically slept more. Their median of 7.5 hours is 1 hour more than the 6.5-hour median for juniors with a first-period class. Variability: the spread of sleep is about the same for both groups, a little larger for the first-period group (IQR 1.3 h vs 1.2 h; SD 1.0 h vs 0.9 h). Shape: the no-first-period group looks roughly symmetric (mean 7.4 ≈ median 7.5, equal gaps from min to Q1 and Q3 to max). The first-period group has a longer lower tail and a low outlier: lower fence = 5.8 − 1.5(1.3) = 3.85, and the 3.5-hour student is below it.

(c) z = (9.2 − 7.4)/0.9 = 1.8/0.9 = 2.0. Juan's 9.2 hours of sleep is 2 standard deviations above the mean sleep of juniors with no first-period class.

(d) Get a list of all students in each grade (about 400 per grade). In each grade, number the students 1-400 and use a random number generator to pick 50 different numbers, skipping repeats. Survey those 50 students from each grade, 4 × 50 = 200 total. This is better than an SRS because it guarantees every grade is represented equally and gives a more accurate estimate.

How it scored

2/2a — Correctly called observational, with the reason that no one was assigned a schedule. The response refuses the causal claim and names a plausible confounding variable.
3/3b — Center, variability and shape are all compared in context. The low outlier is backed with the fence calculation, which is strong support.
2/2c — The z-score is computed with the work shown, and it is interpreted in standard-deviation units relative to the right group's mean.
2/3d — The random process can be repeated, and sampling is done separately within each grade, so two points are earned. The advantage is too vague. "More accurate" needs a reason: sleep likely differs by grade, so stratifying lowers the variability of the estimated mean. The benefit point is lost.
Sleep and First-Period Classes · mid-scoring response5/10

(a) It is an observational study because the council did not do anything to the students. But they can conclude that first period causes less sleep, because students with first period slept a whole hour less on average, which is a big difference.

(b) The no-first-period group slept more, since its mean of 7.4 hours is higher than the 6.4-hour mean for students with first period. The SD for first period is 1.0 and the SD for no first period is 0.9. Both distributions are normal.

(c) z = (9.2 − 7.4)/0.9 = 2.0. This means Juan slept 2 hours more than the average student.

(d) Split the school into 9th, 10th, 11th and 12th grade. Number the students in each grade 1-400 and use a random number generator to choose 50 different students from each grade, for 200 total. Stratified sampling is more representative than an SRS.

How it scored

1/2a — Calling the study observational, with a correct reason, earns the identification point. The causal conclusion then contradicts that answer. A large difference in an observational study still cannot show cause, so the justification point is lost.
1/3b — The center comparison is correct and in context. The two SDs are listed side by side but never compared, so no variability point. "Both are normal" cannot be read off summary statistics, and it misses the low outlier at 3.5 hours, so the shape point is lost.
1/2c — The calculation is correct, with work shown. The interpretation treats the z-score as hours. A z of 2.0 means 2 standard deviations (1.8 hours) above the mean of his own group, so the interpretation point is lost.
2/3d — Strata by grade with a random sample inside each, described clearly enough to repeat, earns two points. "More representative" is not an explanation. It needs to say that sleep likely differs by grade and that stratifying reduces the variability of the estimate, so the benefit point is lost.

Where points die

Using the mean and standard deviation to find outliers or draw a box plot, such as checking whether a value lies beyond mean + 2SD. Box plots and the standard outlier rule use quartiles only. The fences are Q1 − 1.5(IQR) and Q3 + 1.5(IQR), so with an IQR of 18 they sit 27 units outside the box, not 18. If the question names the 1.5 × IQR rule, any other method earns zero even when its answer agrees.

Assuming the whiskers always run to the minimum and maximum. In the modified box plot AP uses, a whisker stops at the most extreme value still inside the fence. Anything beyond the fence is drawn as a separate point. A long upper whisker with dots above it means right skew with high outliers, and a long lower whisker means left skew.

Treating every summary statistic as equally sensitive to an outlier. The mean and SD use every value, so one extreme value drags them a long way. The median and IQR barely move. When a group has an outlier or strong skew, compare medians and IQRs, and say the range or SD is inflated by the extreme value. In right-skewed data the mean sits above the median.

Judging how unusual a value is by its raw distance from the center, or reading a z-score as a number of original units. Being 17 points above the median means nothing until you know the spread. Divide by the SD: z = (value − mean)/SD. Then interpret z as "standard deviations above or below the mean" of the correct group. A z of 2.0 for a sleep value means 2 SDs (1.8 hours here), not 2 hours.

Listing statistics for two groups side by side and calling it a comparison, or leaving out context. Every comparison needs a comparison word and the variable's name: "the median sleep for juniors without a first-period class is 1 hour greater." Cover center, variability and shape, and name any outliers. Never call a distribution Normal from summary numbers alone. You need a graph: a histogram or dotplot that is roughly symmetric and single-peaked with no outliers, or a Normal probability plot that is close to straight. Curves, S-shapes or stray points at the ends of a Normal probability plot signal non-Normality.

Drawing a cause-and-effect conclusion from an observational study because the difference looks large. Only random assignment of treatments supports causation. If the groups existed before the study, a confounding variable could explain the difference, however big it is. In design answers, say blocking or stratifying reduces variability from a known source. Blocking and stratifying are not the fix for bias, which comes from flawed data collection: non-random selection, nonresponse, question wording or measurement problems, or no random assignment of treatments.

Score your own response

0 / 10

Be the harshest reader in the room: award a point only when your response does what the line says, on the page — not what you meant to write.