AP Statistics Bell Ringers: 30 Five-Minute Ideas

Why five minutes beats a warm-up worksheet

A warm-up worksheet asks students to copy something they already wrote down last class. A bell ringer asks them to commit to an answer before they know if it's right, and that commitment is what makes the two minutes of resolution afterward worth having.

AP Statistics rewards a narrow set of skills over and over: read a graph, check a condition before you calculate, interpret a number in the context of the problem. None of those show up on a fill-in-the-blank review sheet. All thirty prompts below ask for one of them, need no printing, and take no student accounts to run. Most work off a single sentence on the board or projector.

Exploring one-variable data

These five put a shape, a center, or a spread in front of students and ask them to pick the right number for the situation, not just compute one.

1. Skewed prices. A histogram of 40 home sales in one neighborhood is skewed right, with one 2.1 million dollar sale sitting apart from a cluster that otherwise runs from 290,000 to 460,000 dollars. Should the neighborhood report the mean or the median sale price? Correct answer needs: median, named as the resistant statistic, with a reason that the mean gets pulled toward the outlier and the skew.

2. Same mean, different spread. Two sections of the same class average 82 on a quiz. Section A has a standard deviation of 3, Section B has a standard deviation of 11. A student in each section scored 70. What does that 70 mean in each section? Correct answer needs: Section A's scores are tighter, so a 70 there is about four standard deviations below the mean and unusual; in Section B a 70 is about one standard deviation below the mean and unremarkable. The z-score calculator checks the arithmetic in a few seconds if you want to project it.

3. Same median, different box. Project two boxplots of commute time for two towns with identical medians, but Town A's box is twice as wide as Town B's. Which town has more variable commute times, and which number proves it? Correct answer needs: Town A, because the width of the box is the interquartile range, and the median being equal says nothing about spread. The descriptive statistics sandbox lets you drag points and watch the box change live if a student wants to argue it out.

4. One number, two variable types. Read one line from a survey: students rated the cafeteria food from 1 to 5, and the survey also recorded which lunch period each student has. Which variable is categorical, and could the 1-to-5 rating be argued either way? Correct answer needs: lunch period is categorical; the rating is quantitative by construction but many classes treat it as ordinal categorical in practice, and a defended answer either way should earn credit as long as the reasoning is stated.

5. The fence. A data set has Q1 = 20 and Q3 = 32. A new value of 55 shows up. Is it an outlier by the 1.5 IQR rule? Correct answer needs: IQR = 12, so the upper fence is 32 + 1.5(12) = 50, and 55 sits past it, so yes.

Two-variable categorical data

A two-way table on the board is the fastest way to catch the mixup between a marginal number and a conditional one, and it catches itself in front of the class before it costs points on the exam.

6. Read the margin. A table of 300 students crosses playing a sport (yes or no) with having a part-time job (yes or no): 90 play a sport and work, 60 play a sport and do not work, 50 do not play a sport but work, and 100 do neither. What fraction of all 300 students play a sport? Correct answer needs: 150 out of 300, or 0.5, read straight off the row total.

7. Read the condition. Using the same table, what fraction of the students who have a job also play a sport? Is that the same number as question 6? Correct answer needs: 90 out of the 140 students who work, about 0.643, and a statement that this is a different, conditional number, not the marginal 0.5. See topic 2.1 for the vocabulary if a student mixes the two up.

8. Same rows, no story. If every conditional distribution in a two-way table looks identical across the rows, what does that say about association between the two variables? Correct answer needs: no association, because identical conditional distributions mean the row variable does not change with the column variable. This is the everyday-language setup for the chi-square test that comes later in the year.

9. Read the segments. Project a segmented bar chart of favorite music genre across three grade levels. Ask which genre shows the biggest shift in share from the youngest grade to the oldest. Correct answer needs: a specific genre named, plus a rough percentage-point estimate of the shift, not just 'that one looks different.'

Collecting data: sampling and experiments

Design questions are where students lose points for using the right word in the wrong sentence. These make them say the word out loud before the stakes are a free-response point.

10. Name the method. A principal numbers all 1,200 students and uses a random number generator to pick 100 for a lunch survey. Name the sampling method, and name one thing that can still go wrong even though the selection was random. Correct answer needs: simple random sample, plus a real threat such as nonresponse from students who do not return the survey. The sampling methods practice set has eight more like it.

11. Who got left out. A cafeteria survey is only handed out during fourth-period lunch. Which kind of bias does that risk, and who does it leave out? Correct answer needs: undercoverage, because students in every other lunch period never had a chance to be selected. The distinction from a nonresponse problem is in undercoverage versus nonresponse bias.

12. Two different randoms. A company recruits 60 volunteers, then randomly assigns 30 to a new training video and 30 to the old one. Can the results support a cause-and-effect claim about the video? Can they be generalized to every employee at the company? Correct answer needs: yes to causation, because treatments were randomly assigned; no to generalizing, because the 60 volunteered rather than being randomly selected. The full reasoning behind both halves of that answer is in can you generalize these results, and the two randomizations are told apart in random sampling versus random assignment.

13. Name the rival. Students who eat breakfast score higher on average on a Monday quiz than students who skip it. Name one confounding variable that could explain the gap instead of breakfast itself. Correct answer needs: a variable linked to both eating breakfast and quiz scores, such as how much sleep a student got the night before, stated as a rival explanation rather than just named in isolation.

Probability

Probability bell ringers work best as a single small number students can find on paper in under a minute, then defend out loud.

14. The trick question. Can two events with positive probability be both mutually exclusive and independent? Correct answer needs: no, with the reason: mutually exclusive events have $P(A \text{ and } B) = 0$, but independence requires $P(A \text{ and } B) = P(A)P(B)$, which can only equal zero if one event already has probability zero.

15. The complement. A weather app gives a 35% chance of rain today. What is the probability it does not rain, and which rule gets you there? Correct answer needs: 65%, from the complement rule: $1 - 0.35$.

16. Do not double-count. In a class of 30, 18 take Spanish, 10 take French, and 4 take both. What is the probability a randomly chosen student takes Spanish or French? Correct answer needs: $\frac{18}{30} + \frac{10}{30} - \frac{4}{30} = \frac{24}{30} = 0.8$, using the addition rule and subtracting the overlap exactly once.

17. Two branches, one question. A factory's Machine A makes 60% of parts with a 2% defect rate, and Machine B makes 40% with a 5% defect rate. What is the probability a randomly chosen part is defective? Correct answer needs: $0.6(0.02) + 0.4(0.05) = 0.032$, built from a tree diagram with weighted branches, not an average of the two defect rates. The tree diagram practice set has more of these.

18. Check, don't guess. In a survey, $P(\text{owns a pet}) = 0.6$, $P(\text{has a yard}) = 0.7$, and $P(\text{owns a pet and has a yard}) = 0.42$. Are the two events independent? Correct answer needs: yes, because $0.6 \times 0.7 = 0.42$ matches the given joint probability exactly. The check itself is walked through in how to check if events are independent.

Random variables and sampling distributions

These four move from a single random variable to what happens when you take a mean of many of them, which is the idea the rest of the course depends on.

19. Is it worth five dollars? A raffle ticket costs five dollars. There is a 1-in-200 chance of winning 500 dollars, otherwise nothing. What is the expected value of buying one ticket, and is it a good deal on average? Correct answer needs: expected winnings of $500 \times \frac{1}{200} = 2.50$, a net expected value of $2.50 - 5 = -2.50$, and the conclusion that it is a bad deal on average. The expected value calculator confirms it in one line.

20. Two standard deviations. Adult male height is approximately Normal with a mean of 70 inches and a standard deviation of 3 inches. Using the empirical rule, about what percent of men are between 64 and 76 inches? Correct answer needs: about 95%, because 64 and 76 are each exactly two standard deviations from the mean.

21. Small sample, big skew. A population of household incomes is heavily right-skewed. Will the distribution of sample means from samples of size 4 look Normal? What about samples of size 100? Correct answer needs: size 4 is too small, so the sampling distribution still shows skew; size 100 is large enough for the central limit theorem to make it approximately Normal. The sampling distribution and CLT visualizer makes the transition visible by dragging the sample size slider.

22. All four conditions. A player makes 70% of free throws and shoots 12 in a row, with each shot independent of the last. Is the number of makes a binomial random variable? Name all four conditions and confirm each one against the setup. Correct answer needs: yes, with all four named: a fixed number of trials (12), two outcomes per trial, a constant probability of success (0.7), and independent trials. How to tell if a situation is binomial spells out the same four.

Inference for proportions and means

Inference bell ringers are mostly damage control. These five target the specific sentences students write wrong on every practice exam, and catching the sentence in five minutes is cheaper than losing the point in May.

23. Fix the sentence. A 95% confidence interval for the share of voters supporting a measure is (0.46, 0.54). A student writes: 'There's a 95% chance the true proportion is in this interval.' What is wrong, and what should it say instead? Correct answer needs: the population proportion is a fixed number, not random, so the interval either captures it or it does not; the correct sentence is that the method captures the true proportion in about 95% of intervals built this way. How to interpret a confidence interval for a proportion has the full wording, and the confidence interval coverage simulator shows the 95% as a running tally instead of a sentence.

24. Fix the other sentence. A test of $H_0: p = 0.5$ returns a p-value of 0.03. A student writes: 'There's a 3% chance the null hypothesis is true.' Correct it. Correct answer needs: the p-value is the probability of a result at least this extreme, assuming the null hypothesis is true, not the probability that the null hypothesis itself is true. See what does a p-value mean.

25. Check before you calculate. Before running a one-sample t-test on 8 measurements, what has to be checked, and what happens to the test if a dotplot of the 8 values shows a strong skew and one clear outlier? Correct answer needs: random sampling, the 10% condition if sampling without replacement, and roughly symmetric data with no outliers when the sample is this small; with a strong skew and an outlier at $n = 8$, the t-procedure is not trustworthy and that has to be stated in the conclusion, not skipped. The conditions checking practice set drills exactly this call.

26. Name both errors. A quality-control test has $H_0$: the batch is safe. Describe, in context, what a Type I error looks like, what a Type II error looks like, and say which one is worse here. Correct answer needs: a Type I error is rejecting a batch that is actually safe (a false alarm, good product discarded); a Type II error is failing to catch a batch that is actually unsafe (an unsafe batch ships); most classes agree the second is worse in this context, but the grade is on the reasoning, not the pick. The Type I error, Type II error, and power visualizer shows both regions moving as the true parameter shifts, and the tradeoff itself is in power versus Type II error.

27. Same subjects, twice. A study measures the same 20 patients' blood pressure before and after a new drug. Is this a paired design or a two-sample design, and which standard error formula follows from that? Correct answer needs: paired, because the same 20 patients were measured twice, so the analysis runs a one-sample t-procedure on the 20 differences rather than the two-sample standard error formula. Paired versus two-sample t-test has the full contrast.

Chi-square and regression

The current AP Statistics course dropped inference for the slope of a regression line along with chi-square goodness of fit, so these three stay inside what the redesigned exam actually tests: the chi-square tests for homogeneity and independence, and reading a regression line as a description rather than an inference.

28. Homogeneity or independence? A researcher takes one sample of 200 people and classifies each one by eye color and hair color, to see whether the two are related. Which chi-square test fits: homogeneity or independence? Correct answer needs: independence, because there is a single sample classified on two variables at once; a homogeneity test would compare separate samples from different populations on one variable. Chi-square homogeneity versus independence has the full rule.

29. Find the expected count. A two-way table of 200 employees crosses shift (morning or evening) with calling in sick: 120 work mornings, 80 work evenings, and 30 of the 200 called in sick. If shift and calling in sick were unrelated, what expected count belongs in the 'morning and sick' cell? Correct answer needs: $\frac{120 \times 30}{200} = 18$, from the expected count formula, row total times column total over the grand total. The idea of comparing this expected count to what was actually observed is in observed versus expected counts.

30. Read the slope. A regression predicting exam score from hours studied gives the line $\hat{y} = 52 + 6.5x$. Interpret the slope in one sentence, in context. Correct answer needs: for each additional hour studied, the model predicts about 6.5 more points on the exam, on average, with the word 'predicted' or 'on average' present so it is not read as a guaranteed gain. How to interpret the slope of a regression line is the reference if a student's wording needs a second pass, and the regression calculator checks the line itself.

How to actually run these

Same slot, every day. The routine does more of the work than the content does. A class that knows the first five minutes are always a bell ringer walks in ready for it.

Do not grade them. The moment a bell ringer counts for points, students start optimizing for the point instead of the thinking, and the honest wrong answer, which is the whole reason to run one, disappears.

Always resolve it. An unresolved prompt teaches a class that the first five minutes do not matter. Two minutes of cold-calling one wrong answer and one right one is enough.

Recycle on purpose. Run the marginal-versus-conditional prompt again in the unit on inference for proportions, and the scope-of-inference prompt again right before the exam. The forgetting is the point, and spaced repetition is the fix, which matters more in a cumulative exam than in almost any other AP subject.

Bell ringers are not a substitute for a full lesson. They are the five minutes that keep an earlier unit alive while a new one is being taught, which is exactly what a cumulative exam rewards. Full unit-by-unit study material lives in the AP Statistics course guide, and a broader set of practice problems for any one skill is in practice.