A student wants to estimate the difference between the proportions of students who bike to school at two large high schools. She selects a random sample of 120 students from School A, which has 1,500 students, and a separate random sample of 80 students from School B, which has 1,100 students. In the samples, 15 students at School A and 6 students at School B bike to school. Which condition for constructing a two-sample z-interval for p_A - p_B is NOT met?
- ALarge Counts, because the school B sample has only 6 successesCorrect
- BRandom, because the two samples have different sizes, 120 and 80
- C10%, because 120 + 80 = 200 is more than 10% of School B's 1,100 students
- DNormality, because both sample proportions, 0.125 and 0.075, are below 0.5
Why A is right
The Large Counts condition needs at least 10 successes and at least 10 failures in each sample. School A has 15 successes and 105 failures, but School B has only 6 successes (and 74 failures), so the condition fails for School B. That means the sampling distribution of p-hat_A - p-hat_B may not be approximately normal, so a z-interval could have a true capture rate different from the stated confidence level. The other conditions hold: both samples are random, and each sample is less than 10% of its own school (120 < 150 and 80 < 110).
Why the other 3 choices miss
- B
- The Random condition is about how the samples were chosen, not whether they are the same size. Both samples are random, and a two-proportion interval does not need equal sample sizes.
- C
- The 10% condition is checked for each sample against its own population: 120 is less than 10% of 1,500 (150), and 80 is less than 10% of 1,100 (110). Adding the two samples together and comparing them to one school is a mistake.
- D
- Normality of the sampling distribution depends on the counts of successes and failures (Large Counts), not on whether p-hat is close to 0.5. Small proportions are fine as long as the expected counts are large enough.