01 · Learn the idea
Know the population first
Prepare 20 identical folded slips: 12 marked B and 8 marked G. The known B proportion is 12/20 = 0.60. Mix well and draw 5 slips without replacement for one sample, count B, then return all slips before the next sample. Repeat at least 12 times. Each slip must have an equal chance to be selected.
For a digital version, assign IDs 1–12 to B and 13–20 to G. Use a random tool to select 5 distinct IDs per sample; reset the pool for each sample. Record the method and actual outputs. A student choosing favorite numbers is not a random generator.
02 · Build understanding
Compare samples, not promises
Convert each B count to a proportion and graph the distribution of sample proportions. Individual samples may be quite different from 0.60; random selection gives a fair method, not a guarantee of an exact match in every small sample. Compare the mean of your repeated sample proportions with 0.60 while recognizing simulation variation.
A biased method, such as taking only slips from a visible B pile, systematically favors one category. Separate sampling variation under a fair method from bias in the method. A sample should represent the population relevant to the question, not merely be easy to collect.
03 · Build understanding
Worked example: from a sample to the whole school
This school and its numbers are made up. Pine Hollow Middle School has 600 students. A computer picks 40 of them at random, and 14 of those 40 say they ride the bus. About how many of the 600 students ride the bus?
Step 1: Turn the sample result into a proportion, because a proportion can be moved from a small group to a big one. 14 out of 40 is 14 ÷ 40 = 0.35.
Step 2: Use the same proportion for the whole school. A random sample gives every student the same chance to be picked, so the sample should look a lot like the school. 0.35 × 600 = 210 students.
Step 3: Check that the estimate matches the sample. 210 out of 600 is 210 ÷ 600 = 0.35, the same proportion as 14 out of 40.
Step 4: Say it as an estimate. About 210 students ride the bus. A different random sample of 40 would likely give a slightly different number.
04 · Build understanding
Worked example: five samples from a jar you already know
These results are invented. A jar holds 200 beads, and 80 of them are red, so the true proportion of red is 80 ÷ 200 = 0.40. Five students each take a random sample of 20 beads, count the red ones and put the beads back. The red counts are 7, 9, 8, 6, 10.
Step 1: Change each count to a proportion by dividing by the sample size, 20. You get 0.35, 0.45, 0.40, 0.30 and 0.50.
Step 2: Compare each one with the true 0.40. One sample matches exactly. The other four miss by 0.05 or 0.10. No sample is far off, and every sample was picked the same fair way. Differences like these are called sampling variation.
Step 3: Find the mean of the red counts: 7 + 9 + 8 + 6 + 10 = 40, and 40 ÷ 5 = 8 beads. As a proportion that is 8 ÷ 20 = 0.40, the same as the jar.
Step 4: See what each sample would predict for the whole jar. Multiply each proportion by 200 to get 70, 90, 80, 60 and 100 red beads. The true number is 80. The estimates sit on both sides of it.
Step 5: Check one estimate against its sample. The sample with 6 red beads predicts 60, and 60 ÷ 200 = 0.30, which matches 6 ÷ 20 = 0.30.
05 · Build understanding
Bigger samples wander less
Here are more invented results from the same jar, where the true proportion of red is 0.40. Five random samples of 10 beads have red counts 2, 6, 4, 5, 3. As proportions these are 0.2, 0.6, 0.4, 0.5 and 0.3. The lowest and highest are 0.6 − 0.2 = 0.4 apart.
Five random samples of 50 beads have red counts 19, 21, 20, 22, 18. As proportions these are 0.38, 0.42, 0.40, 0.44 and 0.36. The lowest and highest are only 0.44 − 0.36 = 0.08 apart.
Both sets of samples were chosen fairly, and both center near 0.40. The larger samples stay closer to the true value. A sample of 10 can easily miss by 0.2. In these results, no sample of 50 missed by more than 0.04.
06 · Build understanding
Three mistakes to watch for
Multiplying the count by the population. In the school example, 14 × 600 = 8,400, which is more students than the school has. Divide first to get the proportion 0.35, then multiply by 600.
Expecting every sample to match exactly. A jar sample with 6 red beads out of 20 gives 0.30, not 0.40. That doesn't mean the jar changed or someone cheated. Random samples of the same size bounce around the true value.
Trusting a sample that was not random. Suppose you only ask the 40 students standing in the bus line. All 40 ride the bus, and 40 ÷ 40 = 1, so you would estimate that all 600 students ride it. The arithmetic is fine. The sample is the problem, because students who walk had no chance to be picked.
Apply and explain
Your turn
Run and record the 12 equal-size samples using slips or a digital generator. Plot their B proportions, mark 0.60, and describe variation. Explain why a sample with 4 B slips does not prove the method is biased. Compare the method with deliberately selecting only B slips.
Use paper or explain your thinking aloud before opening the solution.
Check your reasoning
Actual sample results vary; each proportion is the count divided by 5. Four B slips give 0.80 and can occur by chance in a fair sample. Deliberate B selection is biased. The graph and summary should reflect recorded results rather than expected invented data.
- Use a known population and equal sample sizes.
- Keep selection fair and restore the population between samples.
- Record actual simulated results.
- Distinguish random variation from systematic bias.
Ready to practice?
Try the questions, get feedback, and review your reasoning.
Repeated random samples practice →