Simple random vs stratified vs cluster sampling

The three sampling methods at a glance

All three methods use random chance to pick a sample from a population, but they differ in how they split the population and what they draw from those pieces. In a simple random sample (SRS), you do not split at all: you draw $n$ individuals from one list so that every possible group of size $n$ is equally likely. In a stratified random sample, you split the population into non-overlapping groups called strata that are internally similar, then take a separate SRS inside each stratum and combine the results. In a cluster random sample, you split the population into groups called clusters that each resemble the whole population, take an SRS of the clusters, and then measure every individual in the chosen clusters.

One contrast drives the whole topic. Stratified sampling makes homogeneous groups and samples within each of them; cluster sampling makes heterogeneous groups and samples whole groups. A fourth method, systematic sampling, orders the population and picks every $k$-th unit from a random start. This is Unit 1 topic 1.11, and the exam mostly asks you to read a scenario and name the method.

Simple random sampling: every group of size n is equally likely

The definition of an SRS is stricter than it first sounds. In a simple random sample of size $n$, every sample of size $n$ has the same chance of being selected. That is stronger than saying each individual has an equal chance, because it also requires that every combination of individuals is equally likely. Many biased methods still give each person an equal individual chance, so the group-level condition is what makes an SRS special.

To carry one out, you give every individual a number and then use a random mechanism to choose. Two standard procedures are a random number generator and drawing numbered slips of paper from a hat. AP problems usually want to see the mechanism described: assign labels 1 to $N$, where $N$ is the number of individuals in the population, generate random numbers, and take the first $n$ distinct labels.

A related idea is sampling with or without replacement. Sampling without replacement means an individual can be chosen only once, since it is not returned to the population before the next draw. Sampling with replacement returns the individual each time, so the same person could be selected again. Real surveys almost always sample without replacement.

The SRS is the benchmark. The other methods still use random chance, but they add structure that an SRS does not have, so they are not simple random samples even when every individual has the same chance of being picked.

Stratified vs cluster: the distinction the exam wants

This is the pair students mix up most. Both methods divide the population first, but they divide it for opposite reasons and then sample it in opposite ways.

Stratified sampling divides the population into strata based on a shared attribute, such as grade level, age band, or region, and good strata are homogeneous, meaning the people inside one stratum are alike on the trait that matters. You then take an SRS within every stratum and combine them into one sample. Because every stratum is forced to appear, stratified sampling guarantees representation of each group, and when the strata really are internally similar it tends to produce a less variable estimate than an SRS of the same total size. That lower variability, not just fairness, is the statistical payoff.

Cluster sampling divides the population into clusters, and the goal is the reverse: each cluster should be heterogeneous and mirror the whole population, with the clusters similar to one another. You take an SRS of the clusters and then collect data from every individual in the selected clusters. The reason to do this is practical. If a full list of individuals is hard to get or people are spread out, sampling a few intact clusters (a few schools, a few city blocks) is cheaper and easier to run than chasing scattered individuals.

A quick way to keep them straight: strata are similar within, and you sample some of each; clusters are each like the whole, and you sample all of a few. If the scenario samples a fraction of every group, it is stratified. If it picks a few whole groups and measures everyone in them, it is cluster.

Systematic sampling and a comparison table

A systematic random sample lines the population up in some order and selects members using a random starting point and a fixed, periodic interval. For example, to pick 40 shoppers from a stream of arrivals, choose a random start among the first 10 people and then take every 10th person after that. It is easy to run in the field because you never need a full list in advance, though it can go wrong if the ordering itself has a hidden pattern that lines up with the interval.

Here is how the four methods compare on how they split the population, what they draw, and why you would choose them.

Each method uses randomness, so each can give an unbiased picture of the population when it is carried out well. The choice among them depends on the question, the budget, and what list of the population you actually have.

How the AP exam tests sampling methods

Sampling methods are Unit 1, topic 1.11, part of the 20% to 30% of the multiple-choice section that Unit 1 carries. The topic sits under the Collect Data statistical practice, which is 20% to 30% of the multiple-choice questions on its own, so it is worth knowing cold. The two learning objectives are to identify a sampling method from a description (skill 2.A) and to justify why a method is appropriate for a given question (skill 2.B).

The most common question type gives you a short scenario and asks you to name the method. Work it by asking two questions in order: did the study split the population first, and if so, did it sample within every group or take whole groups?

• A method that selects individuals at random from one undivided list, with no fixed interval, is an SRS.
• A method that samples inside every group is stratified.
• A method that takes a few whole groups is cluster.
• A method that walks down an ordered list at a fixed interval is systematic.

Watch for the trap that stratified and systematic samples are not simple random samples, because they do not let every group of size $n$ occur. Also remember that a random method can still be poorly matched to the question, which is why justification (skill 2.B) matters as much as naming.

To see how sampling connects to designing studies, read experiments vs observational studies. To review the vocabulary of a sample statistic estimating a population parameter, see parameter vs statistic. The full unit lives at Unit 1: Exploring One-Variable Data, and the official framework is at AP Central.