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

Sample size

Sample size appears in 40 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 Sample size is tested

Example 1Medium
A researcher took independent random samples of 200 adults from each of two hypothetical towns. In Town 1, 50% of those sampled supported a new park, and in Town 2, 42% did. From these data, she built a 95% confidence interval for p1 - p2, the difference in the population proportions who support the park. Suppose she had instead sampled 800 adults from each town, got the same sample proportions, and built a 99% confidence interval. Compared with the original interval, the new interval would be
  1. AExactly half as wide as the original interval
  2. BAbout two-thirds as wide as the originalCorrect
  3. CAbout one-third as wide as the original interval
  4. DAbout 1.3 times as wide as the original interval

Why B is right

The width is 2 * z* * sqrt(p-hat1(1-p-hat1)/n1 + p-hat2(1-p-hat2)/n2). With the same sample proportions, making each n four times larger divides the standard error by sqrt(4) = 2. With these data the original standard error is sqrt(0.25/200 + 0.2436/200), about 0.0497, so the 95% margin of error is 1.96(0.0497), about 0.097. The new standard error is about 0.0248, and the 99% margin of error is 2.576(0.0248), about 0.064. That makes the ratio (2.576/1.96)(1/2), about 0.66, or roughly two-thirds as wide.
Why the other 3 choices miss
A
Gets the sample-size effect right (four times the n halves the standard error) but ignores that moving from 95% to 99% raises z* from 1.96 to 2.576, which widens the interval by a factor of about 1.31.
C
Assumes the margin of error shrinks in proportion to n instead of sqrt(n). Dividing by 4 and then multiplying by 1.31 gives about 0.33, but quadrupling n only halves the standard error.
D
Accounts for the larger z* at 99% confidence but assumes the standard error stays the same because the sample proportions did not change. The standard error depends on n as well as on p-hat.
Example 2Harder
A researcher wants to construct a 95% confidence interval for the mean with a margin of error no greater than 2 units. The population standard deviation is estimated to be 10. What is the minimum required sample size? (Use z* ≈ 1.96)
  1. An ≥ 10
  2. Bn ≥ 25
  3. Cn ≥ 97Correct
  4. Dn ≥ 196

Why C is right

Margin of error formula: ME = z*(σ/√n). Set ME ≤ 2: 1.96(10/√n) ≤ 2. Solve for n: √n ≥ (1.96 × 10)/2 = 9.8. n ≥ 9.8^2 = 96.04. Round up to n ≥ 97. The confidence interval or hypothesis test for a mean uses the sample statistic and standard error from the setup.
Why the other 3 choices miss
A
n ≥ 10 is far too small. ME = 1.96(10/√10) = 1.96 × 3.16 ≈ 6.19, which exceeds the desired ME of 2.
B
n ≥ 25: ME = 1.96(10/5) = 3.92 > 2. This sample size is insufficient to achieve the desired precision.
D
n ≥ 196 would arise from n = (z*σ/ME)^2 = (1.96 × 10/2)^2 = 9.8^2 = 96.04 — not 196. n = 196 would be obtained if ME = 1, not ME = 2.
Example 3Harder
A sample of n = 64 observations from a population has x-bar = 105 and s = 24. The researcher tests H0: mu = 100 vs. Ha: mu ≠ 100. The t-statistic is t = (105 - 100)/(24/sqrt(64)) = 1.67 with df = 63. The researcher notes that if the sample had been n = 256 (four times as large) with the same x-bar and s, the t-statistic would have been
  1. A0.84 (halved because larger n increases the denominator)
  2. B6.68 (quadrupled because n quadrupled)
  3. C1.67 (unchanged because x-bar and s are the same)
  4. D3.33 (doubled because n quadrupled and SE halved)Correct

Why D is right

With n = 256: SE = 24/sqrt(256) = 24/16 = 1.5 (compared to 24/8 = 3 for n = 64). t = (105 - 100)/1.5 = 5/1.5 = 3.33. Quadrupling n doubles sqrt(n) (from 8 to 16), which halves the SE, which doubles the t-statistic. This illustrates why larger samples can produce significant results for the same effect.
Why the other 3 choices miss
A
Incorrect. A larger n makes SE smaller (not larger), which makes t larger (not smaller). The denominator decreases, increasing the test statistic.
B
Incorrect. Quadrupling n only doubles the t-statistic (because t depends on sqrt(n), and sqrt(4) = 2). It does not quadruple it.
C
Incorrect. The t-statistic depends on n through the standard error. Quadrupling n changes the SE and therefore the t-statistic.

Studied alongside

Other concepts from Inference for Categorical Data: Proportions.