Skip to content
AP Statistics · Inference for Categorical Data: Proportions

Estimators

Estimators appears in 9 reviewed AP Statistics questions and sits in Inference for Categorical Data: Proportions. Below are worked examples showing exactly how the exam tests it, with the answer marked and every distractor explained.

How Estimators is tested

Example 1Medium
A high school has 2,000 students. The student council wants to estimate p, the proportion of all students at the school who support a later start time. Two data collection plans are proposed. Plan I: Select a simple random sample of 150 students from the school roster and ask each selected student the question. Plan II: Post the question on the school website and use the first 400 students who respond. The sample proportion p-hat from each plan will be used to estimate p. Which plan should the council use, and why?
  1. APlan I, because 400 is more than 10% of the 2,000 students, and that makes the p-hat from Plan II a biased estimator of p.
  2. BPlan I, because random selection makes p-hat an unbiased estimator of p; Plan II has less variability but is likely biased by voluntary response.Correct
  3. CNeither plan, because p-hat is an unbiased estimator of p only after the large counts condition is checked.
  4. DPlan II, because with 400 responses p-hat varies less from sample to sample, so it will be closer to p than Plan I.

Why B is right

p-hat is an unbiased estimator of p, meaning the center of its sampling distribution equals p, only when the data come from a random sample. Plan I meets that condition. Plan II is a voluntary response sample: students with strong opinions are more likely to respond, so its p-hat may be systematically too high or too low. Plan II's larger sample does lower the variability of p-hat, but a larger sample does not remove bias, so Plan I gives the better estimator.
Why the other 3 choices miss
A
Misuses the 10% condition. 400/2000 = 20% does break that condition, but the condition only controls whether the formula sqrt(p(1-p)/n) is accurate for the standard deviation. Sampling more than 10% does not make p-hat biased. Plan II's bias comes from voluntary response.
C
Mixes up the conditions: large counts (np >= 10 and n(1-p) >= 10) decides whether the sampling distribution is approximately normal. It has nothing to do with whether p-hat is unbiased, which depends on random sampling.
D
Mixes up variability with bias: a larger sample makes p-hat vary less, but a precise estimate from a biased method still centers on the wrong value, so being less variable does not mean it is closer to p.

Studied alongside

Other concepts from Inference for Categorical Data: Proportions.