finish the Survey Report with a random sample, an inference and a two-grade comparison, and critique the booth cards.
Phase 4: Build
Do Now. The Census Bureau asks about 3.5 million addresses a year, not every home in the country. Read today’s reading, then answer.
Why can a random sample of 3.5 million addresses describe the whole country?
The reading opens over the lesson; close it to come back to exactly where you were.
Our Survey Report has three parts. (1) How we chose a random sample. (2) An inference about the whole school from the sample. (3) A comparison of two groups using center and variability. Today we finish it and then critique every booth card.
Worked example 1: Priya writes the inference. The team drew 40 names at random from a sample list of all 240 seventh graders. Twelve chose a basketball shoot as the new booth: 12/40 = 30%. Priya writes: “We estimate about 0.30 × 240 = 72 seventh graders would choose the basketball shoot. A different random sample might give a slightly different number, so we say about 72.”
Worked example 2: Mateo compares two grades. The team asked 10 randomly chosen students in each grade how many hours they sleep on a school night (sample data). Grade 7: 8, 9, 7, 8, 9, 8, 10, 7, 8, 6. Grade 8: 7, 8, 6, 7, 7, 8, 6, 9, 6, 6. Grade 7 mean = 80 ÷ 10 = 8 hours; median 8. Grade 8 mean = 70 ÷ 10 = 7 hours; median 7. The mean absolute deviation (MAD) is 0.8 hours for each grade. The means differ by 1 hour, which is 1 ÷ 0.8 = 1.25 MADs.
What can we say? Mateo writes: “In our samples, seventh graders slept about 1 hour more than eighth graders. The groups vary by the same amount, and the difference is a little more than one MAD, so it may be real, but it is not the clear 2-MAD gap from lesson 2.11. With samples of only 10, we should check with a bigger random sample.”
Gallery critique. Walk the booth cards with the class criteria. Use sentence starters: “I can check your probability because …” “Your sample space is missing …” “Your experiment and your count disagree by …, which could mean …”
Check the Survey Report with your team before the gallery walk.
Gallery walk. Leave written feedback on two cards.
Leave feedback on two booth cards: one strength, one fix, and one number you checked a second way. Use the class criteria.
Survey team data check: in a random sample of 10 eighth graders, the minutes spent reading last night were 20, 30, 0, 45, 30, 15, 30, 25, 40, 15 (sample data). Find the mean and the median. Then write one sentence comparing them with a grade 7 sample whose mean was 30 minutes.
In 1936, a popular magazine called the Literary Digest tried to predict the U.S. presidential election between Franklin D. Roosevelt and Alf Landon. It mailed about 10 million postcard ballots and got back about 2.3 million. Based on its poll, it predicted that Landon would win with 57.1% of the popular vote.
Roosevelt won in a landslide, carrying 46 of the 48 states. How could a poll of 2 million people be so wrong? The magazine took its names from its own subscriber list, car registration lists, phone lists and club membership lists. During the Great Depression, those lists leaned toward wealthier families who could afford magazines, cars, phones and clubs, so they left out a large part of the country. Also, only the people who chose to mail back a ballot were counted.
The sample was huge, but it was not random, and size could not fix that. (Source: Kyle Siegrist, Random Services, “The Literary Digest Poll,” randomservices.org/random/data/LiteraryDigest.html, CC BY 2.0.)
Think about it: Which matters more for a good survey: a big sample or a random sample? What would you tell a team that surveys only students in the lunch line?
A good report says what we found, how sure we are and why.
Exit ticket: write the one change you will make to your team's card or report because of today's feedback.
Pacing: warm-up and reading 6 minutes, mini lecture 9, case study 6, guided practice 10, gallery walk 7, closure 2.
Likely misconceptions: believing a bigger sample always beats a random one (the case study); treating a 1-hour difference as important without looking at variability; overclaiming from samples of 10.
Supports: MAD calculation table (value, distance from mean); feedback sentence starters on cards. Stretch: find each grade’s range and IQR. Grade 7: range 4, IQR 9 − 7 = 2; grade 8: range 3, IQR 8 − 6 = 2.
Answer guide: warm-up: every address had a chance; names from a hat; 72; 1.25; true; missing TH. Reading: if chosen households skipped, the people left would not represent everyone and the data would become less valid; we survey everyone drawn for the same reason. Reading-minutes task: sum 250, mean 25 minutes; ordered 0, 15, 15, 20, 25, 30, 30, 30, 40, 45, median (25 + 30) ÷ 2 = 27.5 minutes; the grade 7 sample’s mean was 5 minutes higher. MAD check (Mateo): grade 7 distances 0, 1, 1, 0, 1, 0, 2, 1, 0, 2 sum to 8; grade 8 distances 0, 1, 1, 0, 0, 1, 1, 2, 1, 1 sum to 8; 8 ÷ 10 = 0.8 each. Case study: figures are from Kyle Siegrist’s open (CC BY 2.0) Random Services page, The Literary Digest Poll (randomservices.org/random/data/LiteraryDigest.html, opened 2026-11-20): about 10,000,000 postcards mailed, about 2,300,000 returned, names from subscriber, car registration, phone and club lists, a predicted 57.1% for Landon, and Roosevelt carrying 46 of 48 states. The page also names self-selection (who chose to mail a card back) as a second problem. Keep the discussion on sampling, not on the candidates.
U.S. Census Bureau, Top Questions About the American Community Survey · U.S. government work, public domain
The U.S. Census Bureau runs the American Community Survey, or ACS, every year. It asks about many topics, and the Bureau says the answers help communities plan hospitals and schools, support school lunch programs and build bridges. The Census Bureau does not ask every household. Instead, it chooses a sample.
On its page of top questions, the Census Bureau explains how the sample is chosen. Each address is randomly selected through a process of scientific sampling. The bureau samples addresses in every state, the District of Columbia and Puerto Rico. It chooses about 3.5 million addresses each year.
Why not ask everyone? The bureau says it tries to keep the number of households in the sample as small as possible. A smaller sample costs less and asks fewer people to spend time on the survey.
The bureau also explains why a chosen household cannot simply skip the survey. If selected households were let off, the sample data would quickly become less valid. In other words, the sample would no longer be random. The people left would not represent everyone, and the results would be biased.
Source: U.S. Census Bureau, Top Questions About the American Community Survey (census.gov)