Strata in AP Statistics

In AP Statistics, strata are non-overlapping subgroups of a population whose members share a characteristic (like age group or income level); in stratified random sampling, a separate random sample is taken from every stratum so each group is guaranteed representation.

Verified for the 2027 AP Statistics examLast updated June 2026

What are strata?

Strata (singular: stratum) are the groups you split a population into before you sample, where everyone in a group shares some characteristic that matters to your study. Think of slicing a population into layers, all 18-29 year olds in one layer, all 30-45 year olds in the next, and so on. The whole point is that each stratum is homogeneous, meaning the individuals inside it are similar to each other on the variable you care about.

In a stratified random sample, you take a separate simple random sample (SRS) from every stratum. That's the move that makes it stratified. Because you sample from each layer, no group gets skipped, and your estimates get more precise. The variable you use to form the groups (age, grade level, socioeconomic status) is called the stratification variable, and it should be related to the response you're measuring. Stratifying cabbage rows by distance from a river only helps if distance from the river actually affects aphid counts.

Why strata matter in AP® Statistics

Strata live in Topic 3.3 (Random Sampling and Data Collection) in Unit 3: Collecting Data. They directly support two learning objectives. For AP Stats 3.3.A, you have to identify a sampling method from a study description, which means recognizing 'divided into groups, then randomly sampled from each group' as stratified sampling. For AP Stats 3.3.B, you have to explain why a method is or isn't appropriate (DAT-2.D.1), which usually means explaining that strata are homogeneous within and that sampling from every stratum guarantees representation and reduces variability in estimates. This is one of the most reliably tested ideas in Unit 3 because it's easy to confuse with cluster sampling, and the exam knows it.

How strata connect across the course

Cluster Sample (Unit 3)

Clusters are the mirror image of strata. Strata should be similar within and different between; clusters should each be a mini version of the whole population. With strata you sample from every group; with clusters you randomly pick whole groups and take everyone in them.

Simple Random Sample / Random Number Generator (Unit 3)

Stratified sampling isn't a replacement for an SRS, it's an SRS run multiple times. Inside each stratum you still number individuals and use a random number generator (or table) to select them, per DAT-2.C.2. If the within-stratum selection isn't random, the whole design falls apart.

Sampling without replacement (Unit 3)

When you draw your sample within each stratum, each individual can be selected only once, which is sampling without replacement (DAT-2.C.1). This matters later in Unit 5 when independence conditions for sampling distributions get checked with the 10% condition.

Blocking in Experiments (Unit 3)

Strata in sampling and blocks in experimental design are the same instinct applied to different settings. Both group similar units together to control a known source of variability. Stratify when you're sampling; block when you're assigning treatments.

Are strata on the AP® Statistics exam?

Multiple-choice questions hand you a study description and ask you to name the method or justify it. The classic stem looks like one of these: a sociologist splits families by socioeconomic status and randomly samples 50 from each group, or a researcher categorizes voters into age brackets (18-29, 30-45, 46-65, 66+) and randomly selects 100 from each. Your job is to spot 'random sample from EVERY group' and call it stratified, then explain that it guarantees representation from each subgroup. On the FRQ side, the 2025 exam (FRQ Q2) set up a cabbage field with a river 100 meters south, exactly the kind of scenario where stratifying by distance from the river makes sense because rows at different distances may differ in aphid counts. To earn full credit on questions like this, you can't just name the method. You have to connect the stratification variable to the response variable and say why homogeneous strata improve the sample.

Strata vs Clusters

This is the single most common Unit 3 trap. Strata are built to be internally similar (all low-income families together, all seniors together), and you sample from every stratum. Clusters are built to be internally diverse, each one a snapshot of the full population, and you randomly select a few whole clusters and survey everyone in them. Quick check: sample from ALL groups means strata; sample SOME groups entirely means clusters.

Key things to remember about strata

  • Strata are non-overlapping subgroups of a population where individuals within each group share a characteristic related to what you're measuring.

  • In stratified random sampling, you take a separate random sample from every single stratum, not just some of them.

  • Good strata are homogeneous within and different between, which reduces variability and makes estimates more precise.

  • Strata are the opposite of clusters: strata are similar inside, clusters should each look like a mini population.

  • On the exam, justifying stratification means explaining why the stratification variable is related to the response variable, not just naming the method.

  • Sampling within each stratum is still done randomly and without replacement, usually as an SRS.

Frequently asked questions about strata

What are strata in AP Stats?

Strata are non-overlapping groups within a population whose members share a common characteristic, like age bracket or income level. In stratified random sampling, a separate random sample is drawn from each stratum so every group is represented.

What's the difference between strata and clusters?

Strata are designed to be similar within each group, and you sample from every stratum. Clusters are designed to each represent the whole population, and you randomly select a few entire clusters. If the study samples from all groups, it's stratified; if it picks whole groups, it's cluster sampling.

Do you have to take the same number of people from each stratum?

No. Equal sizes (like 100 voters from each age group) are common in exam problems, but the requirement is that some random sample is taken from every stratum. Sample sizes can be proportional to stratum size instead.

Is a stratified sample better than a simple random sample?

Often, yes, but only if the stratification variable is actually related to the response. Per DAT-2.D.1, every method has trade-offs, so stratifying by something irrelevant adds work without reducing variability. When strata are truly homogeneous, stratified samples give more precise estimates than an SRS of the same size.

Are strata the same as blocks?

They're the same idea in different settings. Strata group similar individuals in a sampling design, while blocks group similar experimental units before random assignment to treatments. Use 'strata' for surveys and 'blocks' for experiments on the AP exam.