A stratified random sample divides the population into non-overlapping subgroups (strata) of similar individuals, then takes a separate random sample from each stratum, guaranteeing every group is represented and producing more precise estimates than a simple random sample alone.
A stratified random sample is a two-step sampling design. First, you split the population into strata, which are groups that are similar within themselves but different from each other (urban vs. rural schools, freshmen vs. seniors, traditional vs. online students). Second, you take a separate random sample from each stratum. Every stratum is guaranteed to show up in your data, which is the whole point.
Why bother? Because if a group matters to your research question, leaving it to chance is risky. A plain simple random sample of 100 students might accidentally grab only 3 seniors. Stratifying by grade level locks in representation from every group and reduces sampling variability, so your estimates are more precise. Think of it as taking several mini-SRSs, one inside each subgroup, instead of one big SRS of everyone.
Stratified random sampling is one of the core sampling designs you learn when collecting data, but it comes roaring back in Unit 8 (Topic 8.5) as a condition for inference. Under learning objective AP Stats 8.5.C, the chi-square test for homogeneity requires data collected from a stratified random sample (or a randomized experiment), while the chi-square test for independence requires a simple random sample. That's not a trivia distinction. The sampling design literally tells you which test you're allowed to run. Separate random samples from multiple populations means homogeneity. One random sample from one population, classified by two categorical variables, means independence. If you can spot the stratified design in a problem stem, you've already answered half the question.
Keep studying AP Statistics Unit 9QyrTKQ1AuLyQwom
Chi-Square Test for Homogeneity (Unit 8)
This is the big one. A homogeneity test asks whether the distribution of a categorical variable is the same across several populations, and you can only compare several populations if you sampled from each one separately. That's exactly what a stratified random sample does, so it's the required data-collection condition under AP Stats 8.5.C.
Cluster Sample (Unit 3)
The classic mix-up. Both methods split the population into groups first, but stratified sampling takes some individuals from every group, while cluster sampling takes every individual from some randomly chosen groups. Strata are built to be similar inside; clusters work best when each one is a mini version of the whole population.
Random Sampling (Unit 3)
Stratification doesn't replace randomness, it organizes it. Within each stratum you still need a random selection mechanism. Stratifying without randomizing inside the strata gives you a convenience sample wearing a costume, and it kills your ability to generalize.
10% Condition (Units 6-8)
Whenever you sample without replacement, including within each stratum, you check that the sample is at most 10% of the population so independence of observations is reasonable. The CED lists this check right alongside the stratified-sample requirement for chi-square tests on two-way tables.
Two main jobs. First, identification questions. A stem describes a design in words, like a researcher splitting all high schools into urban and rural groups and then randomly selecting schools from each group, and asks you to name the method. Your tell is groups first, random selection from every group second. Second, conditions and test-selection questions in Unit 8. A question like "Which sampling method is required for a chi-square test for homogeneity?" is testing whether you know the design-to-test mapping from 8.5.C. On FRQs, when you verify conditions for a chi-square test, explicitly state how the data were collected. Writing "independent stratified random samples were taken from each population, so a test for homogeneity is appropriate" earns credit that a vague "the sample is random" does not. Watch for trap stems where stratified data gets paired with an independence test; that's a mismatch the exam loves to probe.
Both start by dividing the population into groups, which is why they get confused. In a stratified random sample, you sample randomly from EVERY group, and the groups (strata) are designed to be homogeneous within. In a cluster sample, you randomly pick a few groups and survey EVERYONE in them, and ideally each cluster is heterogeneous, like a mini population. Quick memory hook: stratified = some from all groups; cluster = all from some groups. Stratifying improves precision; clustering mostly saves time and money.
A stratified random sample divides the population into homogeneous subgroups called strata, then takes a separate random sample from each stratum.
Stratifying guarantees every important subgroup is represented and reduces sampling variability compared to a simple random sample of the same size.
For a chi-square test for homogeneity, the data should come from a stratified random sample or a randomized experiment; a test for independence needs a simple random sample instead (AP Stats 8.5.C).
Stratified means some individuals from all groups; cluster means all individuals from some groups. Don't swap them on the exam.
When sampling without replacement, still check the 10% condition (n ≤ 10% of N) within your samples before running inference.
Spotting a stratified design in a problem stem signals that you're comparing distributions across multiple populations, which points you toward homogeneity, not independence.
It's a sampling method where you divide the population into non-overlapping subgroups (strata) that are similar within themselves, then take a random sample from each stratum. Every group is guaranteed representation, which makes estimates more precise.
Stratified sampling takes a random sample from every group; cluster sampling randomly picks a few groups and surveys everyone in them. Also, strata should be similar within (all seniors, all urban schools), while ideal clusters are diverse within, like mini versions of the whole population.
Only for the chi-square test for homogeneity, which compares a categorical variable's distribution across multiple populations. The data should come from a stratified random sample or a randomized experiment. The chi-square test for independence requires a single simple random sample instead.
No. Stratifying reduces sampling variability and guarantees subgroup representation, but it can't fix nonresponse bias, bad question wording, or undercoverage of people who never made it into your sampling frame. You still need genuine random selection within each stratum.
Homogeneity compares distributions across two or more separate populations, so you need an independent random sample from each population. Stratified sampling delivers exactly that structure, with each stratum acting as one of the populations being compared.