compare results from small and large random samples and explain why larger random samples usually give estimates closer to the population value.
Phase 2: Investigate
Do Now. Sample situation: Rosa asks 5 random students whether they like the new lunch menu, and 4 say yes. Ben asks 50 random students, and 28 say yes. Whose result would you trust more for the whole school? Why?
Which set of random samples will have results that are closer together?
The reading opens over the lesson; close it to come back to exactly where you were.
Small samples wobble. With 10 cards, one extra red card changes the result by 10 percentage points. With 40 cards, one card changes it by only 2.5 points. So small samples jump around more.
Comparing spreads. To see the effect, draw many samples of each size and compare the dot plots. We can describe the spread of each set of sample results by its range: highest minus lowest.
Worked example 1: samples of 10. Sample data: ten teams drew 10 cards each from a bag that is secretly 30% red. Red counts: 1, 3, 5, 2, 4, 3, 3, 2, 4, 3. Step 1: each card is 10%, so the percents are 10, 30, 50, 20, 40, 30, 30, 20, 40, 30. Step 2: range = 50 − 10 = 40 percentage points. Step 3: mean count = 30 ÷ 10 = 3, which is 30%.
Worked example 2: samples of 40. Sample data: ten teams drew 40 cards each from the same bag. Red counts: 11, 12, 13, 12, 10, 14, 12, 11, 13, 12. Step 1: each card is 2.5%, so the percents are 27.5, 30, 32.5, 30, 25, 35, 30, 27.5, 32.5, 30. Step 2: range = 35 − 25 = 10 percentage points. Step 3: mean count = 120 ÷ 10 = 12, and 12/40 = 30%. Step 4: compare: both sets center on 30%, but the 40-card samples spread only a quarter as much.
Bigger is not always better. A large sample only helps if it is random. Recall lesson 2.03: a bigger biased sample repeats its bias. Also, the Census Bureau tries to keep its sample as small as it can to save money and time. Survey scientists choose a size that is big enough, not as big as possible.
For our report. Our school survey will pick a sample size we can manage, such as 40 to 60 students, and say in the report that a larger random sample would give a steadier estimate.
Hands-on. In teams, draw 10 cards from the class bag, record the red count, and return them. Repeat with 40 cards. Post each result on the matching class dot plot: one for n = 10 and one for n = 40. Use the worked examples’ data if time is short.
Decide and explain.
Sample data: a team surveys 8 random students and finds 6 like the new schedule (75%). Another surveys 64 random students and finds 36 do. Find the second percent and explain which estimate you would put in our report.
A national-only NAEP assessment usually has about how many students?
Write a sentence for our report that explains our sample size choice to families.
Bigger random samples are steadier. Next time we move from one group to two groups and compare their centers.
Exit ticket: a friend says, “A sample of 1,000 people at a stadium is better than a random sample of 100 from the whole town.” Do you agree? Explain.
Pacing: Do Now 4 min, reading 7, mini lesson 10, two-size sampling 14, independent practice and exit ticket 5.
Accuracy: uncertainty falling as the number selected rises, about 10,000 students in a national-only assessment, over 2,000 per state, about 140,000 combined, the small-sample comparison outcome and the cause-and-effect warning are from the NAEP page opened 2026-11-20. Checked: 10-card counts sum 30, mean 3 = 30%, range 10%-50% = 40 points; 40-card counts sum 120, mean 12 = 30%, range 25%-35% = 10 points; 11/40 = 27.5%; 28/50 = 56%; 36/64 = 56.25%.
Likely misconceptions: “bigger sample = no bias”; confusing percent with percentage points.
Supports: pre-drawn axes for the two dot plots, one above the other with the same scale. Stretch: about how many cards would make one card worth 1 percentage point? (100.)
Answer guide: 36/64 = 56.25%, about 56%; the larger sample is steadier. Stadium sample: disagree; it is large but not random for the town (it over-represents sports fans).
National Center for Education Statistics, NAEP: Statistical Significance and Sample Size · U.S. government work, public domain
NAEP chooses students at random, so its results include some uncertainty from sampling. The page explains that in a good sampling design, this uncertainty gets smaller as the number of students selected gets larger. Results for large groups tend to have smaller standard errors than results for small groups.
The page gives real sizes. A national-only NAEP assessment usually has about 10,000 students. Some assessments have separate samples of over 2,000 students in each state, and these are combined. Together they make samples of about 140,000 students, so the national results from them have much smaller standard errors.
Sample size also affects comparisons. When NAEP compares the average scores of two groups, one possible outcome is that the sample does not allow a reliable test. The page says this can happen when the sample for a group is small.
The page also warns about what NAEP cannot do. Its comparisons cannot prove cause and effect. Many things can affect how students do, such as school policies, resources and who the students are.
Source: National Center for Education Statistics, NAEP: Statistical Significance and Sample Size (nces.ed.gov)