U3.1 Random Sampling Methods
Master AP Statistics random sampling: distinguish observational studies from experiments and apply SRS, stratified, cluster, and systematic sampling correctly.
What you'll do in this lesson
A voice-first session with the Crimsora tutor on U3.1 Random Sampling Methods, then targeted practice and FRQs — with the tutor adapting to where you get stuck.
What this lesson covers
Before you can trust a conclusion drawn from data, you have to know how the data were collected. In Unit 3, everything begins with two big questions: was this an observational study or an experiment, and how were the individuals selected? Getting these right determines whether you can talk about associations, cause and effect, or generalizing to a population.
This lesson teaches you to separate observational studies from experiments and to identify and apply the four random sampling methods the AP exam loves to test: simple random sampling, stratified, cluster, and systematic. Nail these definitions now, because later topics on bias, experimental design, and inference all build directly on them.
This lesson teaches you to separate observational studies from experiments and to identify and apply the four random sampling methods the AP exam loves to test: simple random sampling, stratified, cluster, and systematic. Nail these definitions now, because later topics on bias, experimental design, and inference all build directly on them.
Observational Study vs. Experiment
The single most important distinction in data collection is whether researchers impose a treatment or simply observe what already happens.
In an observational study, researchers measure variables without trying to influence the response. They might survey students about sleep and grades, but they do not assign anyone to sleep more or less. Because groups form naturally, lurking variables can differ between them, so an observational study can show association but not causation.
In an experiment, researchers deliberately apply treatments to subjects and then compare responses. Because treatments are assigned (ideally at random), an experiment can support cause-and-effect conclusions.
A common exam trap: students see the word "random" and assume an experiment. But random sampling (who gets studied) is different from random assignment (which treatment each subject gets). A study can randomly sample people and still be observational if no treatment is applied. When you read a scenario, ask directly: did anyone decide what treatment each subject received? If yes, it is an experiment; if the researcher only recorded existing conditions, it is observational.
In an observational study, researchers measure variables without trying to influence the response. They might survey students about sleep and grades, but they do not assign anyone to sleep more or less. Because groups form naturally, lurking variables can differ between them, so an observational study can show association but not causation.
In an experiment, researchers deliberately apply treatments to subjects and then compare responses. Because treatments are assigned (ideally at random), an experiment can support cause-and-effect conclusions.
| Feature | Observational Study | Experiment |
|---|---|---|
| Treatment imposed? | No | Yes |
| Can show causation? | No, only association | Yes, if well designed |
| Key tool | Random sampling | Random assignment |
| Main threat | Confounding | Lack of control |
Simple Random Sampling (SRS) and Systematic Sampling
A simple random sample (SRS) of size is chosen so that every possible group of individuals has an equal chance of being selected. That is a stronger condition than each individual having an equal chance. You can generate an SRS by numbering everyone in the population and using a random number generator or table to pick distinct numbers until you have individuals.
The defining property of an SRS is that any combination of members is equally likely. This is the benchmark all inference procedures assume, so the exam treats SRS as the gold standard.
A systematic sample selects every th individual after a random starting point. For example, to sample 50 of 1000 people, choose a random start between 1 and 20, then take every 20th person. Systematic sampling is easy to carry out and often gives good results, but it is not an SRS because not every group of size is possible—once the starting point is chosen, the entire sample is fixed.
A frequent misconception: "everyone has an equal chance" guarantees an SRS. Systematic sampling can give every individual an equal chance while still failing the SRS definition, because most possible subsets can never be selected. On the AP exam, be precise about whether the method satisfies the all-subsets-equally-likely condition.
The defining property of an SRS is that any combination of members is equally likely. This is the benchmark all inference procedures assume, so the exam treats SRS as the gold standard.
A systematic sample selects every th individual after a random starting point. For example, to sample 50 of 1000 people, choose a random start between 1 and 20, then take every 20th person. Systematic sampling is easy to carry out and often gives good results, but it is not an SRS because not every group of size is possible—once the starting point is chosen, the entire sample is fixed.
A frequent misconception: "everyone has an equal chance" guarantees an SRS. Systematic sampling can give every individual an equal chance while still failing the SRS definition, because most possible subsets can never be selected. On the AP exam, be precise about whether the method satisfies the all-subsets-equally-likely condition.
Stratified and Cluster Sampling
Both stratified and cluster sampling divide the population into groups first, but they use those groups in opposite ways.
In stratified sampling, you split the population into strata of similar individuals (for example, by grade level), then take a separate random sample from every stratum. The goal is to reduce variability by ensuring each subgroup is represented. Strata should be homogeneous within and different from each other.
In cluster sampling, you split the population into clusters that ideally each resemble the whole population (often by geography, like classrooms or city blocks), then randomly select entire clusters and sample everyone inside the chosen clusters. Clusters should be heterogeneous within, each a mini-version of the population.
The classic mix-up: students swap these two. Remember that stratified sampling reaches into all groups but takes only part of each, while cluster sampling picks some groups but takes all of them. Stratified sampling generally reduces sampling variability; cluster sampling is used mainly for convenience and cost. On free-response questions, you must both name the method and describe the correct steps to earn full credit.
In stratified sampling, you split the population into strata of similar individuals (for example, by grade level), then take a separate random sample from every stratum. The goal is to reduce variability by ensuring each subgroup is represented. Strata should be homogeneous within and different from each other.
In cluster sampling, you split the population into clusters that ideally each resemble the whole population (often by geography, like classrooms or city blocks), then randomly select entire clusters and sample everyone inside the chosen clusters. Clusters should be heterogeneous within, each a mini-version of the population.
| Method | How groups are used | Ideal group composition |
|---|---|---|
| Stratified | Sample from every group | Homogeneous within, different between |
| Cluster | Sample all of some groups | Each cluster mirrors population |
How the Exam Tests Sampling Methods
AP questions rarely just ask for a definition. Instead, they describe a real scenario and ask you to identify the method, explain why it fits, or design a procedure.
When asked to identify a method, look for signals: dividing into similar groups and sampling each means stratified; dividing into location-based groups and taking whole groups means cluster; every th item means systematic; a random number generator picking individuals means SRS.
When asked to describe how to carry out a method, use complete, repeatable language. For an SRS, say: label each individual to , use a random number generator to produce distinct integers, ignoring repeats, until you have individuals, and select those labeled. Vague answers like "pick randomly" lose points.
A subtle point the exam rewards: explaining the purpose. Stratified sampling is chosen to guarantee representation and reduce variability; cluster sampling is chosen for practicality when the population is spread out. Connecting method to purpose shows statistical reasoning.
Finally, remember that random sampling lets you generalize to the population, while random assignment lets you establish cause. Keep those two ideas in separate mental boxes—the upcoming lessons on bias, design, and inference depend on you not confusing them.
When asked to identify a method, look for signals: dividing into similar groups and sampling each means stratified; dividing into location-based groups and taking whole groups means cluster; every th item means systematic; a random number generator picking individuals means SRS.
When asked to describe how to carry out a method, use complete, repeatable language. For an SRS, say: label each individual to , use a random number generator to produce distinct integers, ignoring repeats, until you have individuals, and select those labeled. Vague answers like "pick randomly" lose points.
A subtle point the exam rewards: explaining the purpose. Stratified sampling is chosen to guarantee representation and reduce variability; cluster sampling is chosen for practicality when the population is spread out. Connecting method to purpose shows statistical reasoning.
Finally, remember that random sampling lets you generalize to the population, while random assignment lets you establish cause. Keep those two ideas in separate mental boxes—the upcoming lessons on bias, design, and inference depend on you not confusing them.
Key terms
- Observational Study.
- A study in which researchers measure variables without imposing any treatment, so it can reveal association but not causation.
- Experiment.
- A study in which researchers deliberately impose treatments on subjects and compare responses, allowing cause-and-effect conclusions when well designed.
- Simple Random Sample (SRS).
- A sample of size chosen so that every possible group of individuals is equally likely to be selected.
- Stratified Sampling.
- Dividing the population into homogeneous strata and taking a separate random sample from each stratum to ensure representation and reduce variability.
- Cluster Sampling.
- Dividing the population into clusters that each resemble the whole population, then randomly selecting entire clusters and sampling all members within them.
- Systematic Sampling.
- Selecting every th individual from a list after a randomly chosen starting point.
- Random Assignment.
- Using chance to allocate subjects to treatment groups in an experiment, which balances lurking variables and permits causal conclusions.
Worked example
A high school has 1,200 students divided evenly across four grade levels. The principal wants to survey 80 students about a proposed schedule change and wants to be sure each grade is represented in proportion. Identify an appropriate sampling method, describe how to carry it out, and state whether this is an observational study or an experiment.
First decide the study type. The principal is surveying students about their opinions, not assigning any treatment. No treatment is imposed, so this is an observational study, and results can show association or describe opinions but cannot establish cause and effect.
Next choose a method. The principal wants each grade represented in proportion, which is exactly the purpose of stratified sampling. The four grade levels are natural strata because students within a grade tend to be more alike, and the goal is guaranteed representation.
Now describe the procedure. Each grade has students, and 80 total means students per grade. Within each grade, label the 300 students to . Use a random number generator to produce distinct integers from 1 to 300, ignoring repeats, until 20 are chosen, and select those students. Repeat separately for all four grades. This yields 20 students from each grade for a total of 80.
Finally, justify why stratified fits better than an SRS here: a single SRS of 80 could, by chance, over- or under-represent a grade, while stratifying guarantees proportional representation and reduces variability between grades.
Next choose a method. The principal wants each grade represented in proportion, which is exactly the purpose of stratified sampling. The four grade levels are natural strata because students within a grade tend to be more alike, and the goal is guaranteed representation.
Now describe the procedure. Each grade has students, and 80 total means students per grade. Within each grade, label the 300 students to . Use a random number generator to produce distinct integers from 1 to 300, ignoring repeats, until 20 are chosen, and select those students. Repeat separately for all four grades. This yields 20 students from each grade for a total of 80.
Finally, justify why stratified fits better than an SRS here: a single SRS of 80 could, by chance, over- or under-represent a grade, while stratifying guarantees proportional representation and reduces variability between grades.
Practice questions
A researcher lists all 500 members of a gym alphabetically, randomly picks one of the first 10 members, and then surveys every 10th member after that. Which sampling method is this?
- Simple random sample
- Stratified sample
- Cluster sample
- Systematic sample
Answer: Systematic sample
Choosing a random starting point and then selecting every 10th individual is the definition of systematic sampling. It is not an SRS because once the start is fixed, most possible groups of members can never occur—only individuals spaced 10 apart are selected together.
A city is divided into 40 neighborhoods, each containing a broad mix of residents. Researchers randomly select 5 neighborhoods and interview every resident in those 5. Explain which sampling method this is, why the neighborhood composition matters, and whether cause-and-effect conclusions are possible.
Answer: This is cluster sampling; conclusions are associational, not causal.
Selecting whole groups (neighborhoods) rather than sampling within every group makes this cluster sampling. It works best when each cluster is heterogeneous—a mini-version of the whole city—so that a few clusters still represent the population. Because no treatment is imposed, this is an observational study, so it can describe or show association but cannot establish cause and effect.
Explain the key difference between stratified sampling and cluster sampling, and give the ideal within-group composition for each.
Answer: Stratified samples from every group and needs homogeneous strata; cluster samples entire selected groups and needs heterogeneous clusters.
In stratified sampling you take a random sample from each of the groups, so strata should be homogeneous within and different between one another, which reduces variability. In cluster sampling you randomly choose some groups and include everyone in them, so each cluster should ideally resemble the whole population. The distinction between sampling within all groups versus taking all of some groups is what the exam checks.
FAQ
- What is the difference between random sampling and random assignment?
- Random sampling is how you choose who is in your study, and it lets you generalize results to a population. Random assignment is how you allocate subjects to treatments in an experiment, and it lets you conclude cause and effect. A study can use one, both, or neither.
- Why is systematic sampling not considered a simple random sample?
- An SRS requires that every possible group of size is equally likely. In systematic sampling, once you pick the random starting point, the whole sample is determined, so most possible subsets can never be selected—even though each individual may have an equal chance of inclusion.
- How do I remember stratified versus cluster sampling?
- Think 'stratified = sample from all groups, part of each' and 'cluster = sample all of some groups.' Strata are made of similar individuals; clusters are each meant to look like the whole population.
- Can an observational study ever prove causation?
- No. Because no treatment is imposed, groups may differ due to confounding variables, so an observational study can only establish association. Only a well-designed experiment with random assignment can support cause-and-effect conclusions.
Learn this with a teacher, not a page
The Crimsora tutor teaches U3.1 Random Sampling Methods live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.