AP Statistics: how to study for the exam (the content in our words, the four-step inference answer that scores, the calculator discipline, and a plan)

AP Statistics is less about computing and more about saying what a computation means. You describe data, design a study, compute a probability, and then reach the inference topics, where a confidence interval or a test is set up, carried out and read back in the context of the question. Most free-response points are earned by the four-step shape of an inference answer, and most are lost by skipping a step. Here is the content in our words, the four steps with one problem worked, what the calculator is for, and an eight-week plan.

The AP Statistics units, in our words

The College Board revised AP Statistics for the 2026–27 school year, with the first revised exam in May 2027, and regrouped the content. If your study guide lists nine units, it describes the older grouping. The ideas below are the same either way; check the current course and exam description for the exact unit names and for which inference procedures it keeps.

The content, in the order it builds:

1. Describing one variable. Shape, centre and spread; histograms, box plots, the mean, median and standard deviation. Describing data by hand covers the computations.
2. Describing two variables. Scatter plots, correlation, the least-squares line and residuals.
3. Collecting data. Sampling methods, experiments against observational studies, random assignment, blocking, and the sources of bias.
4. Probability and random variables. The rules of probability, expected value, and the binomial and geometric distributions.
5. Sampling distributions. How a statistic varies from sample to sample. Every inference topic rests on this one. The normal distribution and z-scores is the tool it uses.
6. Inference. Confidence intervals and significance tests for proportions, for means, and for categorical data with chi-square tests.

The inference topics look like several units, but they are one idea applied to different kinds of question: estimate a parameter, or test a claim about it, and say what the result means. Confidence intervals, p-values and hypothesis tests explains each sentence of a result.

The exam in one table

Described in our words. The revision changed some counts, so the College Board's exam page is the only place to read the current question numbers and timings.

Section	What it holds	Weight
I, multiple choice	Questions across the whole course	Half the score
II, free response	Longer questions, several of them inference, written out in full	Half the score

The two halves are combined into a score from 1 to 5. A graphing calculator and a formula and table sheet have long been part of the exam; read the College Board's current policy on both before you rely on them. No AP question is reproduced on this page.

The four-step inference answer

Every AP Stats FRQ on inference wants the same shape. In our words:

1. State. Name the parameter in context ("p is the proportion of all students at this school who…"). Give the hypotheses, or say what the interval is for.
2. Plan. Name the procedure ("a one-proportion z test"). Check each condition with evidence from the question. A condition asserted without evidence earns nothing.
3. Do. Show the statistic, the standard error, and the test statistic or interval, with the formula and numbers written in.
4. Conclude. Give the decision at the stated level, and read the p-value or interval back in the words of the question, naming the population.

The fault that costs most is a correct number with no context sentence. "Reject H₀" is not a conclusion. "There is convincing evidence that the proportion of all students at the school who bring lunch differs from one half" is.

Practise on the released free-response questions on AP Central, under time, and self-score against the published scoring guidelines. They are not reproduced here.

Worked example: one proportion, a test and an interval, done the FRQ way

This problem is ours, written for this page. It is not an AP item.

A school claims that half its students bring lunch from home. In a random sample of 200 students, 120 bring lunch. Is there evidence that the proportion differs from one half?

State

p is the proportion of all students at the school who bring lunch from home.

H₀: p = 0.5

Hₐ: p ≠ 0.5

Plan

A one-proportion z test.

• Random: the sample was random (given).
• Large counts: 200 × 0.5 = 100 expected successes and 100 expected failures, both at least 10.
• Independence: 200 is assumed to be less than 10 % of the school's students. The question does not say, so write the assumption down.

Do

Sample proportion: p̂ = 120 ÷ 200 = 0.6

Standard error under H₀: SE = √(0.5 × 0.5 ÷ 200) = √0.00125 ≈ 0.0354

Test statistic: z = (0.6 − 0.5) ÷ 0.0354 ≈ 2.83

Two-sided p-value: 2 × P(Z > 2.83) ≈ 0.0047

Conclude

The p-value, about 0.0047, is below 0.05. At the 5 % level, there is convincing evidence that the proportion of all students at the school who bring lunch from home differs from one half.

The interval

For an interval, the standard error uses p̂, not the claimed value.

SE = √(0.6 × 0.4 ÷ 200) = √0.0012 ≈ 0.0346

Margin of error: 1.96 × 0.0346 ≈ 0.068

Interval: 0.6 ± 0.068, so about (0.532, 0.668)

Read back: we are 95 % confident that the interval from 0.532 to 0.668 captures the true proportion of all students at the school who bring lunch from home. The interval does not contain 0.5, which agrees with the test.

The common wrong line is "95 % of students are in this interval." The interval is about one number, the proportion of the whole school. It does not describe individual students, and it is not "a 95 % chance the true value is in it" either: the 95 % describes the method, which captures the parameter in 95 % of samples.

Checkable figures: p̂ = 0.6; SE ≈ 0.0354; z ≈ 2.83; p ≈ 0.0047; interval ≈ (0.532, 0.668).

The calculator discipline (and the AP Statistics formula sheet)

The calculator runs the procedure. The paper earns the points.

• Write the formula with the numbers in before the calculator's answer. If you press a wrong key, the reader can still see a correct method and give credit for it. A bare number with no work shown gives a reader nothing to credit.
• Know which outputs the question wants. A calculator's test screen gives a dozen numbers. The answer is the test statistic, the p-value and a sentence. Copying the whole screen is not an answer.
• Name the procedure in words, not by the calculator's menu label.
• Use the formula sheet as a prompt for the "plan" step. It reminds you which standard error goes with which procedure. It does not tell you which procedure the question needs. That is the skill being tested.

Where MDM4U overlaps, for Canadian readers

MDM4U, Ontario's Grade 12 data management course, covers a good part of AP Statistics' early content. MDM4U's map has the whole course. In our MDM4U course map, the overlap looks like this:

AP Statistics content	In MDM4U	New in AP
Describing one and two variables	Yes: central tendency, spread, quartiles, scatter plots, correlation, the line of best fit	Residual plots read in more depth
Collecting data	Yes: types of data, sampling, experiments and observational studies	Blocking and the language of design, tested hard
Probability and random variables	Yes: counting, probability, expected value, binomial and geometric distributions	Little
Sampling distributions	Partly: the normal distribution, z-scores, the normal approximation	Sampling distributions as their own topic
Inference	Partly: estimating a proportion and the margin of error	Significance tests, inference for means, chi-square

An MDM4U student starts AP Statistics strong on description and probability, and needs the inference topics and the four-step habit from the beginning.

How to study for AP Statistics: the method

• Procedures by worked example, then cold. Sweller and Cooper (1985) found that students who studied worked examples learned algebra procedures faster than students who solved the same problems unaided. Study one worked test or interval, then do the next one with the example closed. Worked examples, then cold has the routine.
• The four steps on every released FRQ, under time. Self-score against the guidelines, step by step.
• Mixed sets from week 4. The exam never says which procedure a question needs. Rohrer and Taylor (2007) found that practice shuffled across problem types tested better than practice blocked by type, even though it felt harder. Mix proportions, means and chi-square in one set. Why mixing beats blocking explains it.
• An error log by step, not by unit. Each miss is a state, plan, do or conclude miss. Most students find that their misses cluster in plan (conditions) and conclude (context). The practice-test method and the error log is the system.
• Vocabulary cards for the design topics, each with your own example: random assignment against random sampling, blocking, stratifying, confounding.

The eight-week plan

Week	Worked examples, then cold	Mixed set	Free response
1	Diagnostic: one released exam under time, the log by step; describing one and two variables	None yet	The diagnostic's, self-scored
2	Collecting data; probability and random variables	Weeks 1–2	One released question
3	Sampling distributions	Weeks 1–3	One released question
4	Inference for proportions and means; the four steps on two questions	Proportions and means together	Two inference questions
5	Chi-square and the remaining inference topics	All inference procedures mixed	Two inference questions
6	A second released exam; the top three steps from the log	The log's top three	The exam's, self-scored
7	Review by step	Full mixed sets	Two full free-response sets under time
8	The final week: light review, hardest topics early	One short set a day	None after day 3

Week 8 follows chapter 9 of the free playbook, the seven-day exam plan: the log's last weak steps early, a light final day, and a full night's sleep. Roediger and Karpicke (2006) found that practising retrieval beat rereading on a test a week later, which is why every week carries a timed free-response question rather than a reread of the notes.

AP Statistics, on one screen

• The content: describing data, collecting data, probability, sampling distributions, then inference; revised for 2026–27, so check the current unit names.
• The exam: multiple choice and free response, each half the score, 1 to 5; current counts on the College Board's page.
• The four steps: state, plan, do, conclude, with context in the first and the last.
• The interval's meaning: it captures the parameter, never the students.
• The calculator rule: formula and numbers on paper first, then the calculator.
• The plan: worked examples then cold, mixed inference sets from week 4, a second released exam in week 6, a light final week.