AP Statistics Unit 5 ReviewSampling Distributions

AP Statistics Unit 5, Sampling Distributions, is about what happens when you take a statistic (like a sample mean or sample proportion) and ask how it would vary if you repeated your sample over and over. The single biggest idea is the Central Limit Theorem, which says that with a large enough sample size, the distribution of sample means is approximately normal even when the population isn't. This unit is 7-12% of the AP exam, and it's the bridge that turns probability (Unit 4) into inference (Units 6-9).

What this unit covers

Why your sample isn't like mine (and the normal model that handles it)

• Two people taking random samples from the same population get different statistics. That's sampling variability, and it's the whole reason this unit exists. The variation may be random (just chance) or non-random (a flaw in how data was collected).
• The normal distribution gets a comeback tour from Unit 1, but now you treat it as a continuous random variable. The area under the curve over an interval IS the probability that a value lands in that interval.
• You'll calculate normal probabilities in both directions. Given a value, find the probability (using z-scores, a table, or a calculator's normalcdf). Given a probability or percentile, find the boundary value (invNorm). For example, "the lowest 10% of values" means finding $x_a$ where $P(X < x_a) = 0.10$.
• You also have to judge when a normal model is appropriate. Normal curves are symmetric and bell-shaped, so they only approximate distributions that share those features.

The sampling distribution idea and the Central Limit Theorem

• A sampling distribution of a statistic is the distribution of that statistic's values across all possible samples of a given size from a population. Picture taking thousands of samples, computing $\bar{x}$ each time, and making a histogram of those means. That histogram is the sampling distribution.
• The Central Limit Theorem (CLT) says that if sample values are independent and n is sufficiently large (n ≥ 30 is the standard cutoff for means), the sampling distribution of the sample mean is approximately normal, no matter what shape the population has. Skewed population, bimodal population, doesn't matter.
• You can also estimate sampling distributions using simulation, which is how the CLT usually gets demonstrated. A randomization distribution is built the same way, by repeatedly reshuffling and recomputing a statistic.

Point estimates and bias

• A sample statistic is a point estimator of the corresponding population parameter. $\bar{x}$ estimates $\mu$, and $\hat{p}$ estimates $p$.
• An estimator is unbiased if, on average across all possible samples, its value equals the population parameter. Bias is about the center of the sampling distribution, not about any single sample being "wrong."
• Even unbiased estimators vary from sample to sample. That variability can be modeled with probability, which is what the standard deviation formulas in this unit do.

The four sampling distributions you must know

• Sample proportion $\hat{p}$ has mean $\mu_{\hat{p}} = p$ and standard deviation $\sigma_{\hat{p}} = \sqrt{p(1-p)/n}$. It's approximately normal when $np \geq 10$ and $n(1-p) \geq 10$ (the Large Counts condition).
• Sample mean $\bar{x}$ has mean $\mu_{\bar{x}} = \mu$ and standard deviation $\sigma_{\bar{x}} = \sigma/\sqrt{n}$. It's normal if the population is normal, or approximately normal when $n \geq 30$ by the CLT.
• Difference in sample proportions $\hat{p}_1 - \hat{p}_2$ has mean $p_1 - p_2$ and standard deviation $\sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}$, with the Large Counts check applied to both samples.
• Difference in sample means $\bar{x}_1 - \bar{x}_2$ has mean $\mu_1 - \mu_2$ and standard deviation $\sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}$. Normal if both populations are normal, approximately normal if both sample sizes are at least 30.
• One fine-print rule covers all four. The standard deviation formulas assume sampling with replacement (independence). When sampling without replacement, the true standard deviation is a bit smaller, but if the sample is less than 10% of the population, the difference is negligible. That's the 10% condition.

Unit 5, Sampling Distributions at a glance

Why Unit 5, Sampling Distributions matters in AP Stats

This is the unit where AP Stats pivots from describing data to making conclusions about populations. Every confidence interval and significance test in the rest of the course is just a sampling distribution wearing a different outfit. If you understand why $\sigma/\sqrt{n}$ shrinks as n grows, the second half of the course makes sense instead of feeling like a pile of formulas.

• It answers the course's central question, which is how confident you can be in a conclusion when variation is always present.
• Probabilistic reasoning lets you anticipate patterns. You can't predict one sample's mean, but you can predict how thousands of sample means will behave.
• The conditions you check here (random, 10% condition, Large Counts or normality) are the exact same conditions you'll write out on every inference problem for the rest of the year.

How this unit connects across the course

• The normal distribution, z-scores, and the empirical rule from one-variable data (Unit 1) come back here, except now you apply them to distributions of statistics instead of distributions of individual values.
• Random sampling methods (Unit 3) are what make sampling distributions trustworthy. A biased sampling method produces a biased estimator, and no amount of CLT magic fixes that.
• Random variables, expected value, and combining independent variables (Unit 4) supply the machinery. The formulas for differences in this unit come straight from the rules for variances of sums and differences of independent random variables.
• Everything here pays off immediately in inference for proportions (Unit 6) and means (Unit 7), where the standard deviations from this unit become standard errors inside confidence intervals and test statistics. The same logic extends to chi-square (Unit 8) and slopes (Unit 9).

Key formulas and procedures

• $z = \frac{x - \mu}{\sigma}$ standardizes a value so you can find normal probabilities; for a statistic, replace x, $\mu$, and $\sigma$ with the statistic, its mean, and its standard deviation.
• $\mu_{\hat{p}} = p$ and $\sigma_{\hat{p}} = \sqrt{\frac{p(1-p)}{n}}$ give the center and spread of the sampling distribution of a sample proportion.
• $\mu_{\bar{x}} = \mu$ and $\sigma_{\bar{x}} = \frac{\sigma}{\sqrt{n}}$ give the center and spread of the sampling distribution of a sample mean.
• $\sigma_{\hat{p}_1 - \hat{p}_2} = \sqrt{\frac{p_1(1-p_1)}{n_1} + \frac{p_2(1-p_2)}{n_2}}$ is the standard deviation for a difference in proportions from two independent samples.
• $\sigma_{\bar{x}_1 - \bar{x}_2} = \sqrt{\frac{\sigma_1^2}{n_1} + \frac{\sigma_2^2}{n_2}}$ is the standard deviation for a difference in means. Notice you always add inside the radical, even for a difference.
• Large Counts condition ($np \geq 10$ and $n(1-p) \geq 10$) verifies approximate normality for proportions; check it for both groups in two-sample problems.
• CLT condition ($n \geq 30$, or a normal population) verifies approximate normality for means.
• 10% condition (sample size less than 10% of population) lets you use the standard deviation formulas when sampling without replacement.
• Procedure for any probability question about a statistic: state the mean and standard deviation of the sampling distribution, verify the shape is approximately normal, compute a z-score, find the area, and interpret in context.

Unit 5, Sampling Distributions on the AP exam

Unit 5 is 7-12% of the AP exam. On the multiple-choice section, expect questions that ask you to compute a probability involving $\bar{x}$ or $\hat{p}$, identify the correct mean and standard deviation of a sampling distribution, recognize what happens to the spread when n changes (quadrupling n cuts the standard deviation in half), and pick out which scenario satisfies the conditions for approximate normality. Conceptual CLT questions are common, like identifying which histogram could be a sampling distribution of means from a skewed population.

On free-response questions, sampling distribution work usually appears inside a larger problem. You might calculate $P(\bar{x} > \text{some value})$ as one part of a question, justify why a sampling distribution is approximately normal by naming and checking conditions, or explain in context what a probability means. Sampling distributions also anchor investigative-task-style questions that build a new statistic and ask you to reason about its distribution from a simulation. Wherever it shows up, full credit requires three moves: correct parameters, a stated and verified shape condition, and an interpretation tied to the specific context. "There is about a 4% chance that a random sample of 50 batteries has a mean lifetime above 510 hours" earns points; a bare number doesn't.

Essential questions

• Why do statistics vary from sample to sample, and how can that variation itself be predictable?
• How does the Central Limit Theorem let us use the normal model even when we know nothing about the population's shape?
• What makes an estimator trustworthy, and how do bias and variability describe different kinds of "wrong"?
• How does sample size change what we can conclude from a sample?

Key terms to know

• Sampling distribution: The distribution of values of a statistic across all possible samples of a given size from a population.
• Central Limit Theorem (CLT): With independent observations and a large enough sample size, the sampling distribution of the sample mean is approximately normal regardless of the population's shape.
• Parameter: A number describing a population, like $\mu$ or $p$; usually unknown.
• Statistic: A number computed from a sample, like $\bar{x}$ or $\hat{p}$; it estimates a parameter.
• Point estimator: A statistic used as a single-value estimate of a population parameter.
• Unbiased estimator: An estimator whose average value across all possible samples equals the population parameter.
• Sampling variability: The natural sample-to-sample variation in a statistic's value.
• Standard deviation of a statistic: A measure of how far a statistic typically falls from the parameter; it shrinks as sample size grows.
• Large Counts condition: The check $np \geq 10$ and $n(1-p) \geq 10$ that justifies a normal approximation for sample proportions.
• 10% condition: When sampling without replacement, the sample should be less than 10% of the population so the standard deviation formulas still work.
• Continuous random variable: A variable that can take any value in an interval, with probabilities given by areas under a density curve.
• z-score: The number of standard deviations a value sits from the mean, used to find normal probabilities.
• Randomization distribution: A simulated distribution of a statistic built by repeatedly re-randomizing data, used to estimate sampling distributions.

Common mix-ups

• The CLT is about the SHAPE of the sampling distribution of the mean. The facts that $\mu_{\bar{x}} = \mu$ and $\sigma_{\bar{x}} = \sigma/\sqrt{n}$ are true for any sample size; the CLT only adds approximate normality when n is large.
• Don't confuse the population distribution, the distribution of one sample's data, and the sampling distribution of a statistic. The first two can be any shape; only the third gets squeezed toward normal as n grows.
• For differences of independent statistics, variances ADD even though you're subtracting. The standard deviation formula always has a plus sign under the square root.
• Bias and variability are separate problems. A high-variability estimator can still be unbiased (it's centered correctly, just noisy), and a low-variability estimator can be badly biased (consistent, but consistently wrong).