"Lab day," Comet says, shaking a cloth bag. "20 tokens. 12 are marked M for morning."
"So the bag's true proportion is 0.6," Wren says. "Today we pretend we do not know that. We sample."
"Draw 10, one at a time, putting each back," Comet says. "Count the Ms. Then do it again."
Nova logs ten rounds: 6, 7, 5, 8, 6, 7, 6, 4, 7, 6 Ms out of 10.
"Round 2 gave 7 of 10, that is 70%," Wren says. "The rule's margin for n = 10 is 29%. What do you notice?"
"Huge," Comet says. "70% plus or minus 29% covers almost everything."
"Ten draws is a tiny sample," Wren says. "The bracket is honest about that."
Nova glows. "Would you like a hint? Every round's interval should still catch the true 60%."
The crew's 10 rounds are made up for the Run Log. Yours will differ, and the comparison is the point.
| Round | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 |
|---|---|---|---|---|---|---|---|---|---|---|
| Ms in 10 draws | 6 | 7 | 5 | 8 | 6 | 7 | 6 | 4 | 7 | 6 |
The crew's mean count is 6.2, near the true 6. Each round's sample percent wandered, and each bracket was wide.
With n = 10, the margin for a 70% sample is 29%. A tiny sample gives an honest but almost useless bracket.
| Statement | True or false? |
|---|---|
| 12 ÷ 20 = 0.6 | ? |
| A sample of 10 gives a smaller margin than a sample of 50. | ? |
| Every one of the crew's 10 rounds gave exactly 6 Ms. | ? |
| A sample percent can miss the true percent and still be a fair estimate. | ? |
Good lab work. Tomorrow the same idea estimates a mean: the crew's stretch times and a bracket in seconds.