What the exam is
This is the AP math course where you write more than you compute. The whole year is one argument about uncertainty: you have a sample, and you want to say something honest about the population behind it. Statistics gives you the tools to say it and to state how wrong you might be.
Our bank splits the course into four blocks, weighted the way real released questions weight them. Exploring Data is 26 percent: shape, center, spread, outliers, z-scores, boxplots, least-squares regression, residual plots, r and r-squared. Collecting Data is 14 percent: random sampling, bias, experimental design, blocking, and confounding. Probability, Random Variables, and Sampling Distributions is 20 percent: conditional probability, independence, binomial and geometric models, expected value, and the Central Limit Theorem. Statistical Inference is 40 percent: confidence intervals and significance tests for proportions, means, chi-square counts, and regression slope. College Board organizes the same material into nine units and publishes a weight range for each unit rather than a single number. Each of our four blocks falls inside those published ranges, and the 40 percent we give Statistical Inference sits at the high end of its range.
That 40 percent figure hides how much inference leans on everything before it. A confidence interval is built directly from a sampling distribution: take your point estimate, then add and subtract a critical value times the standard error, where the standard error is the estimated spread of that sampling distribution. A test is only valid if the data were collected randomly, and it only supports a cause-and-effect claim if subjects were randomly assigned to treatments. Students who treat the first three blocks as warm-up get punished in the fourth.
One idea from Collecting Data governs the entire course, so learn it early. Random selection of subjects is what lets you generalize a result to a larger population. Random assignment of subjects to treatments is what lets you claim one variable caused a change in another. These are separate design features doing separate jobs. An observational study with a beautiful random sample still cannot establish causation, and a tightly randomized experiment on volunteers still cannot be generalized past people like those volunteers.
The prerequisite is second-year algebra. There is no calculus. Most students take it in junior or senior year, frequently alongside precalculus, and it is the AP math course most often chosen by students heading toward biology, psychology, economics, nursing, or business.
The format
Three hours, split into two 90-minute halves that are worth 50 percent each.
Section I is 40 multiple-choice questions with five options, A through E. That is two minutes fifteen seconds per question. Nothing is deducted for a wrong answer, so never leave a blank. Most items are one step: read a boxplot, name the correct procedure, interpret a p-value, identify the confounding variable, decide whether a result generalizes. Only a handful need real calculator work.
Section II is six free-response questions. Questions 1 through 5 are short answer at roughly 13 minutes each, and each is worth 7.5 percent of your total score. Question 6 is the Investigative Task. It gets about 25 minutes and is worth 12.5 percent of your total score, which is 25 percent of Section II. That makes it worth more than any single short-answer question at 7.5 percent, but less than any two of them combined at 15 percent. Budget your time accordingly: do not let Q6 eat two short questions. It drops you into an unfamiliar setup and asks you to stretch a familiar idea to fit it. You are not supposed to recognize it. You are supposed to reason out loud. Earlier parts of that question usually set up the last part, so answer them in order and reuse your own results.
Each free-response question is scored holistically from 0 to 4: complete, substantial, developing, minimal, no credit. Readers first mark every part of the question E for essentially correct, P for partially correct, or I for incorrect, then combine those letters into the 0-to-4 score. Two E marks and a P typically produce a 3. This is why a partially finished answer written in full context outscores a correct number sitting alone on the page.
A graphing calculator with statistical capability is allowed on both sections. You are handed a formula sheet plus z, t, and chi-square tables. The formula sheet does not tell you which formula applies, which is the part actually being tested.
AP Statistics is administered as a hybrid digital exam. You answer the multiple-choice section in Bluebook on a device, and you write your free-response answers by hand in a paper booklet. In Section I, no one reads your work; only the option you select is scored, so keep scratch work fast and rough on the paper you are given. In Section II, showing work is the score: name the procedure, write each condition next to the evidence that it holds, show the statistic you computed, and finish with a conclusion in context. Confirm the current year's format with your teacher, but expect to hand-write all free-response work.
How to study for it
Study backward from inference, because that is where 40 percent of the points are and where the other blocks get cashed in.
Start by building a one-page procedure chooser. Every inference question funnels into one of about eight procedures, and picking the wrong one costs you the entire question no matter how clean your arithmetic is. Sort by three questions. Is the variable categorical or quantitative? One sample or two? If two, are they paired or independent? That routes you to a one-proportion or two-proportion z-procedure, a one-sample, paired, or two-sample t-procedure, chi-square goodness-of-fit, chi-square for independence, chi-square for homogeneity, or a t-test for the slope of a regression line. Independence and homogeneity are separate procedures in the CED, and they share their mechanics: same expected-count formula, same test statistic. What separates them is how the data were collected. Independence uses one sample cross-classified on two variables. Homogeneity compares several independent samples on one variable. Free response asks you to justify which one applies, so describe the sampling design in your answer rather than just naming the test.
Next, attach conditions to each row of that page. Random data collection every time. The 10 percent condition when you sample without replacement. Large Counts for proportions, and the check changes with the procedure: for a significance test use the null value, so n times p-naught and n times (1 minus p-naught) must both be at least 10, while for a confidence interval you use the sample proportion, so n times p-hat and n times (1 minus p-hat) must both be at least 10. Normality or a large enough sample for means. All expected counts at least 5 for chi-square. Write the condition, then write the evidence that it holds. Readers give credit for checking, not for claiming.
Then drill the four-step frame until it is automatic: state the parameter and hypotheses, name the procedure and verify conditions, compute the statistic and p-value, conclude in context. The conclusion sentence must compare the p-value to alpha, make a decision about the null hypothesis, and name the actual variable and population. A conclusion with no context is a P, not an E.
Use your calculator as a computer, never as an answer. A screen full of output with no named distribution and no parameters earns nothing. Write what you asked it to find.
Grade your own free-response work against published scoring guidelines using E, P, and I. Doing this on ten released questions teaches you more than fifty multiple-choice items, because it shows you exactly which sentence you keep forgetting to write.
Finally, rehearse the interpretations that get graded constantly and studied rarely. Slope in context. What r-squared means. What a residual plot with curvature tells you. The difference between interpreting a 95 percent confidence interval and interpreting the 95 percent confidence level. What a p-value is, and that it is not the probability the null hypothesis is true. A p-value assumes the null is true and measures how unusual your sample would be under that assumption.
Budget a little time for the two error types as well. A Type I error rejects a true null, and its probability is alpha, the significance level you set in advance. A Type II error fails to reject a false null. Power is the chance of detecting a real effect, and it rises with a bigger sample, a bigger true effect, a larger alpha, and a smaller standard deviation in the population or response. Questions here are usually about naming the error in the context of the story and stating its real-world consequence.
A timeline that works
Eight weeks out, take a full released exam under real timing and score it yourself. Use the result to sort the four blocks into strong and weak, then rebuild your weak blocks from the bottom. Work through one inference procedure per study session: hypotheses, conditions, formula, calculator steps, conclusion sentence. Write each one onto your procedure chooser page as you finish it. If Exploring Data is your weak spot, spend the sessions on reading displays and on regression output tables, since those appear in both sections.
Four weeks out, switch from learning to selecting. Do mixed sets where the procedures are shuffled, so you have to decide what kind of problem you are looking at before you do anything. Add free-response questions in timed pairs, then score them with the official rubric. Track which letters you lose. Almost everyone loses the same two: an unchecked condition and a conclusion with no context. Also start timing your multiple-choice pace at 20 questions in 45 minutes.
One week out, stop learning new material. Read your procedure page daily until you can reproduce it blank. Redo the free-response questions you scored lowest on, not the ones you enjoyed. Do one final full-length timed exam, ideally at the same hour as your real test. Confirm your calculator works and has fresh batteries. You are allowed two graphing calculators, and they must have statistical capability. You do not have to clear the memory, and stored programs and stored formulas are permitted.
The night before, do nothing heavy. Skim your procedure page and your list of interpretation sentences once. Pack two calculators if you can, plus pencils and photo ID. Sleep. On exam day, answer every multiple-choice question even if you are guessing, and on free response write a sentence for every part, because a partial answer scores and a blank never does.
The units
In teaching order. Heaviest is Exploring One-Variable Data and Collecting Data at 25% of the exam — if you are short on time, that is where the points are.
Exploring One-Variable Data and Collecting Data
627 questions · 151 key terms
Probability, Random Variables, and Probability Distributions
209 questions · 59 key terms
Inference for Categorical Data: Proportions
317 questions · 67 key terms
Inference for Quantitative Data: Means
206 questions · 45 key terms
Regression Analysis
120 questions · 32 key terms
How to work through it
Read the unit guide
Each unit page shows what the exam asks for, the terms it draws on, and worked examples from easier to harder.
Drill the terms you blanked on
Every key term has its own page with the questions that test it — 354 of them.
Practice under pressure
Open practice runs on the free set. One purchase unlocks the full 1,479-question bank and progress tracking.
Common questions
What is on the AP Statistics exam?
AP Statistics is organized into 5 units. The heaviest is Exploring One-Variable Data and Collecting Data at about 25% of the exam. The multiple-choice section runs 42 questions in 90 minutes. Each unit page in this library lists the concepts that unit draws on and shows worked examples.
Which AP Statistics unit should I study first?
If you are revising with limited time, start with Exploring One-Variable Data and Collecting Data — it carries roughly 25% of the exam, more than any other unit. The unit list on this page shows every unit's share, so you can work down from the heaviest.
Is this AP Statistics material free?
The study guides, the 354-concept index, and 209 worked examples are free with no account. The full 1,479-question bank, adaptive practice, and progress tracking are part of the one-time AP purchase, which covers every AP course.
Where do these questions come from?
They are written in-house and then put through a review pipeline before they can appear here: the answer key is re-solved in a separate review pass, the explanation is checked for leaked reasoning, sourcing and attribution are verified, and any copyrighted or unverifiable quoted text is removed. Only items that pass every gate are published — 1,479 for AP Statistics.
How many practice questions do I get for AP Statistics?
209 are open to everyone with no account, spread so that every unit and every indexed concept has at least one worked example. The remaining 1,270 come with the AP purchase.