"Lab day," Comet says, setting the bag on the picnic table. "Draw a token, write it down, put it back, shake."
"Forty draws," Wren says. "I sort each one two ways. Even or odd. Low or high, with lane 5 counted as high."
Comet draws. "Six." Wren marks even and high. "Three." Odd and low. The tally grows.
After forty, Wren reads the four boxes. "9 even and low. 11 even and high. 12 odd and low. 8 odd and high."
Nova projects the table with totals. "Would you like a hint? Compare P(even given low) with P(even)."
"Given low, 9 of 21," Comet says. "3/7. The bag says 1/2. Close, not equal."
"What do you notice?" Wren asks. "Forty draws wobble. The bag does not owe us an exact match."
The crew's counts are made up for the Run Log. Yours will differ, and the comparison is the point.
| low (1 to 4) | high (5 to 8) | Total | |
|---|---|---|---|
| even | 9 | 11 | 20 |
| odd | 12 | 8 | 20 |
| Total | 21 | 19 | 40 |
In the crew's run, P(even and low) = 9/40 and P(even) × P(low) = 21/80. In the bag itself, both are exactly 1/4.
A sample of forty draws lands near the true values, not on them. Tomorrow you learn the exact test for independence.
| Statement | True or false? |
|---|---|
| 9 + 11 + 12 + 8 = 40 | ? |
| Putting the token back keeps every draw at the same eight equally likely outcomes. | ? |
| Forty draws will always match the bag's true probabilities exactly. | ? |
| P(even given low) uses the low column as its total. | ? |
Good lab work. Tomorrow you test independence exactly and see why a conditional probability uses the group total.