Sample size (n) is the number of observations collected from a population. On AP Statistics, n drives nearly everything in inference. It shrinks the margin of error (width is proportional to 1/√n), increases the power of a test, and shows up in every condition check (n > 30, n ≤ 10% of N, large counts).
Sample size, written as n, is simply how many individuals or observations you collect from a population. But in AP Stats, n is not just a count. It's the single number that controls how trustworthy your inference is.
Three big things happen as n grows. First, your estimates get more precise. The standard error of a sample mean is s/√n and of a sample proportion is √(p̂(1-p̂)/n), so a bigger n means a smaller standard error and a narrower confidence interval. The width of an interval is proportional to 1/√n, which means you need to quadruple n to cut the width in half. Second, the Central Limit Theorem kicks in. When n is sufficiently large (the usual benchmark is n > 30 for means), the sampling distribution of x̄ is approximately normal even if the population is skewed. Third, your significance tests get more powerful. Larger samples make it easier to detect a real effect, which lowers the probability of a Type II error. One important catch, though. A huge sample cannot fix a biased sampling method. If your data came from a voluntary response sample, collecting more of it just gives you a more precise wrong answer.
Sample size is the connective tissue of the entire second half of AP Stats, Units 5 through 9. The CED names it explicitly and repeatedly. AP Stats 7.3.C and 6.3.C ask you to identify the relationship between sample size, interval width, confidence level, and margin of error (width is proportional to 1/√n). AP Stats 9.3.C extends the same idea to slopes, where the interval for the slope of a regression model tends to narrow as n increases. AP Stats 6.7.C lists increasing sample size as a factor that decreases the probability of a Type II error, which means it raises power. And AP Stats 5.3.A makes n the trigger for the Central Limit Theorem.
Beyond those direct objectives, n appears in every single condition check you'll write on the exam. The 10% condition (n ≤ 10% of N) for independence, the n > 30 normality check for means, the np ≥ 10 and n(1-p) ≥ 10 large counts checks for proportions, and expected counts greater than 5 for chi-square tests all hinge on sample size. If you can explain what n does and verify the conditions it appears in, you've covered a huge slice of what inference FRQs grade.
Keep studying AP Statistics Unit 8
Margin of Error (Units 6-7)
Margin of error is where sample size does its most visible work. Since ME = critical value × standard error, and every standard error has a √n in the denominator, a bigger n directly shrinks the margin of error. The CED even has you rearrange the ME formula to solve for the minimum n needed to hit a target margin of error (AP Stats 6.2.C).
Central Limit Theorem (Unit 5)
The CLT is a promise that activates at a certain sample size. When n is sufficiently large (usually n > 30), the sampling distribution of the sample mean is approximately normal no matter what shape the population has. That's the entire reason the n > 30 condition exists in Units 7 and beyond.
Power of a Test (Unit 6)
Per AP Stats 6.7.C, increasing sample size decreases the probability of a Type II error, which means power goes up. The intuition is that more data shrinks the standard error, so a real difference between the truth and the null hypothesis becomes easier to spot. When an FRQ asks how to increase power without changing α, 'increase the sample size' is the go-to answer.
10% Condition (Units 5-9)
This is the one place where a bigger sample can hurt you. When sampling without replacement, n must be at most 10% of the population size N so observations can be treated as independent. Sample size also sets expected counts in chi-square tests, where expected count = (sample size)(null proportion) and all expected counts should exceed 5 (AP Stats 8.2.D, 8.2.E).
Sample size gets tested in three recurring ways. First, conceptual MCQs ask what happens to interval width, margin of error, or power when n changes. Know the punchline that width is proportional to 1/√n, so quadrupling n halves the width. Second, every inference FRQ requires condition checks built on n. The 2023 FRQ on omega-3 supplements gave you 19 patients, a small sample where you can't lean on the CLT, and the 2024 FRQ about an exercise center with thousands of members is exactly where the 10% condition matters. Multi-group problems like the 2021 soft-drink survey across independent samples make you verify large counts for each sample separately. Third, calculation questions ask you to solve for the minimum n needed to achieve a given margin of error by rearranging ME = z*√(p̂(1-p̂)/n). One more exam habit to build. If a question describes a flawed sampling method, do not say 'they should have collected more data.' Bias does not shrink as n grows, and graders specifically reward recognizing that.
Sample size n is how many observations you collect; population size N is how many individuals exist in the whole group you're studying. Students often assume bigger populations demand bigger samples, but precision depends almost entirely on n, not N. A sample of 1,000 estimates a population of 50,000 about as well as it estimates 300 million. N only matters for the 10% condition, where you check n ≤ 10% of N to justify treating observations as independent when sampling without replacement.
The width of a confidence interval is proportional to 1/√n, so to cut the margin of error in half you must multiply the sample size by four.
Increasing sample size decreases the probability of a Type II error, which means larger samples give a test more power to detect a false null hypothesis.
The Central Limit Theorem says the sampling distribution of x̄ is approximately normal when n is sufficiently large, with n > 30 as the standard benchmark when the population is skewed.
Sample size appears in condition checks everywhere, including n ≤ 10% of N for independence, np ≥ 10 and n(1-p) ≥ 10 for proportions, and expected counts above 5 for chi-square tests.
You can find the minimum sample size needed for a target margin of error by rearranging the formula ME = z*√(p̂(1-p̂)/n) and solving for n.
A larger sample size cannot fix a biased sampling method; it just produces a more precise estimate of the wrong thing.
Sample size, n, is the number of observations collected from a population. In AP Stats it determines the precision of estimates (standard errors shrink by √n), whether the Central Limit Theorem applies (n > 30 for skewed data), and the power of significance tests.
No. Bias comes from how the data were collected, not how much was collected. A voluntary response sample of 10,000 is still biased; a larger n just gives you a more precise estimate of the wrong value. This is a classic AP trap answer.
Sample size n is the number of observations you actually collect; population size N is the size of the entire group of interest. Precision depends on n, not N. The only place N shows up on the exam is the 10% condition, where you verify n ≤ 10% of N.
Because n sits inside a square root in every standard error formula, like s/√n for a mean. Interval width is proportional to 1/√n, and √4 = 2, so multiplying n by 4 divides the width by 2. The CED states this directly in topics 6.3 and 7.3.
Yes. Per AP Stats 6.7.C, increasing sample size decreases the probability of a Type II error, and since power = 1 − P(Type II error), power goes up. It's the cleanest way to boost power without raising the significance level α.