AP Statistics Survival Guide: How to Answer FRQs for Full Marks

Quick answer: AP Statistics free-response questions are scored part by part as essentially correct, partially correct or incorrect, then combined into a 0-4 score. Five recycled ideas carry most of those rubric points: the three distributions, checking conditions, the four-move hypothesis test, error types in context, and scope of inference.

Honest disclosure first: I sit IGCSE and A-Levels, not AP, so I have no AP Statistics score of my own to wave around. What I have is a pile of released College Board scoring guidelines and three friends who took the exam and let me compare what they actually wrote against what the rubric wanted. The gap between what they knew and what they scored was almost never mathematical. It was wording.

How AP Statistics free-response is actually scored

Each FRQ is worth 0 to 4 points, and readers do not tick boxes and add up. They rate each lettered part as essentially correct (1 point), partially correct (half a point) or incorrect (zero), then combine those into a whole-number score, rounding holistically if you land between two.

The shape of the exam matters. Section I is 40 multiple-choice questions in 90 minutes, half your score. Section II is six free-response questions in 90 minutes, the other half, with the last being the Investigative Task. That is around 15 minutes per FRQ, and the difference between a partial and an essentially correct is usually one sentence of context — not one more calculation.

So the honest version of "five concepts that earn most of the marks" is this: these five ideas are what the rubric language keeps asking you to demonstrate. Nobody can promise you a score. But if you cannot do these five in full sentences, you will leak half-points everywhere.

Idea 1: population, sample and sampling distributions are three different things

They sound identical and are not. The population distribution is the spread of the real quantity across everyone. The sample distribution is the spread of values in your one actual sample, the thing you can draw a histogram of. The sampling distribution is the spread of a statistic — a mean or proportion — across every possible sample of size n.

That third one is theoretical and it is the one inference runs on. When the Central Limit Theorem says the sampling distribution of the sample mean becomes approximately Normal with standard deviation sigma / sqrt(n), it is saying nothing about the population and nothing about your single sample.

The drill: every time you read a stats question, say out loud which of the three it is about. If a question says "describe the shape of the distribution of sample means", writing about the raw data is an incorrect, not a partial. To get the distinction hammered in with fresh examples, throw the phrase at our explainer.

Idea 2: conditions before calculations, always

State and check the conditions before you compute anything, in writing, with the numbers filled in. Readers award points for the checking, not for knowing the checks exist.

Worked example. A random sample of 250 students at a large school is taken, and 160 say they use a calculator app daily. Build a 95 percent confidence interval for the true proportion.

1. Random: stated in the question — the sample was randomly selected.
2. Large counts: 250 times 0.64 = 160 successes and 250 times 0.36 = 90 failures, both at least 10.
3. Independence: 250 students is less than 10 percent of all students at a large school.
4. p-hat = 160/250 = 0.64. Standard error = sqrt(0.64 times 0.36 / 250) = sqrt(0.0009216) = 0.0304.
5. Interval = 0.64 plus or minus 1.96 times 0.0304 = 0.64 plus or minus 0.0595, so (0.580, 0.700).
6. Interpretation: we are 95 percent confident that the interval from 0.580 to 0.700 captures the true proportion of all students at this school who use a calculator app daily.

Notice step 6. "There is a 95 percent chance the true proportion is in this interval" is wrong — the parameter is fixed, the interval is what varies. That single swap turns an essentially correct into a partial.

Idea 3: the four-move hypothesis test

Every significance test on the paper is the same four moves: hypotheses, conditions, test statistic with p-value, conclusion in context linked back to alpha.

Worked example. A teacher claims her class revises for more than 90 minutes a night on average. A random sample of 36 students gives a mean of 97.5 minutes with standard deviation 21 minutes. Test at the 5 percent level.

1. H0: mu = 90 minutes. Ha: mu is greater than 90 minutes, where mu is the true mean nightly revision time for the population of students.
2. Conditions: random sample, stated. n = 36 is at least 30, so the sampling distribution of the mean is approximately Normal by the Central Limit Theorem. 36 is under 10 percent of the population.
3. t = (97.5 - 90) / (21 / sqrt(36)) = 7.5 / 3.5 = 2.14 with 35 degrees of freedom, giving a p-value of about 0.020.
4. Since 0.020 is less than alpha = 0.05, we reject H0. We have convincing evidence that the true mean nightly revision time for these students is greater than 90 minutes.

Move 4 is where marks disappear. "Reject H0" alone is a partial. You need the comparison to alpha, the decision, and a context sentence naming the actual variable. Defining mu in words back in move 1 is another half-point most students skip.

Idea 4: Type I, Type II and power, always in context

Naming the error type is worth nothing on its own. The rubric wants the consequence, in the situation described.

• Type I error: rejecting a true null. False positive.
• Type II error: failing to reject a false null. False negative.
• Power = 1 minus the probability of a Type II error. Power rises when n rises, when alpha rises, and when the true effect is further from the null.

Contextualised answer for a vaccine trial: "A Type I error would mean concluding the vaccine is effective when in fact it is not, so an ineffective vaccine would be distributed and people would believe they were protected." That is essentially correct. "Rejecting H0 when H0 is true" is a partial, every time. Generate a few scenarios in the quiz tool and write the consequence sentence for each.

Idea 5: scope of inference — sampling versus assignment

The one my friends said they were least ready for, and a two-by-two you can memorise in five minutes.

• Random sample and random assignment: generalise to the population and claim cause and effect.
• Random sample only: generalise, but no causal claim.
• Random assignment only: causal claim within the group studied, but no generalising.
• Neither: no causal claim, no generalising.

Any FRQ asking "can we conclude X causes Y" or "does this apply to all students" is testing this square. Answer both halves separately and name the design feature justifying each.

Test yourself

1. A 90 percent confidence interval for a mean is (12.4, 15.6). Write the interpretation sentence that a reader would score as essentially correct.
2. n = 40 and p-hat = 0.15. Do the large-counts conditions hold? Show both numbers.
3. Researchers randomly assign volunteers who signed up online to two exercise plans. What can and cannot be concluded, and why?

See what your section scores add up to with the AP Statistics score calculator.

FAQ

Is AP Statistics harder than AP Calculus?

Different hard. The arithmetic is much lighter, but the writing load is far heavier — you are marked on explanation quality, not just answers. Students strong at algebra and weak at written justification often find it more frustrating than calculus. If the calculation side worries you, the math solver walks through standard-error arithmetic step by step.

How many free-response questions are on the AP Statistics exam?

Six, in 90 minutes, worth half the exam. The sixth is the Investigative Task, which asks you to extend a method you were not taught directly. Budget about 12 minutes each for questions 1 to 5 and leave 25 to 30 minutes for the last one.

Do I lose marks for skipping conditions if my calculation is right?

Yes. Conditions are their own rubric component on inference questions, so a perfect interval with unchecked conditions caps that part at partial. Write them out with the actual numbers substituted, not just the names.

Can I still get the point if my answer is numerically wrong?

Often, yes. Rubrics reward correct method and correct interpretation of your own value, so a wrong number carried consistently through a correct procedure with a contextual conclusion still earns credit. Show every step and never leave a bare number. Timed practice in the mock exam builds the habit fastest.

In short

AP Statistics rewards students who write like statisticians, not students who calculate fastest. Get the three distributions straight, check conditions in writing, run the four-move test, contextualise your errors, and answer scope questions in two halves. Those five habits are where the rubric points live.