Review AP Statistics Unit 3 to build the foundation for every inference procedure in the course. This unit covers how data are collected through sampling methods and experimental designs, and why those choices determine whether conclusions about populations or causes are valid.
Use the topic guides, key terms, and practice questions available for this unit to work through each sampling method and experimental design concept.
Unit 3 asks a deceptively simple question: can you trust the data? The answer depends entirely on how the data were collected. This unit covers two parallel tracks: how to sample from a population and how to design an experiment. Each track has its own logic, vocabulary, and set of conclusions you are and are not allowed to draw.
Random sampling methods use chance to select individuals from a population. A simple random sample gives every group of a given size an equal chance of selection. Stratified sampling divides the population into homogeneous strata and samples within each. Cluster sampling selects whole groups at random. Systematic sampling uses a random start and fixed interval. Each method has trade-offs in cost, precision, and feasibility.
An experiment imposes treatments on experimental units and measures a response variable. A well-designed experiment includes comparison of at least two groups, random assignment of treatments, replication across multiple units, and control of confounding variables. Completely randomized, randomized block, and matched pairs designs differ in how they handle variability among units.
Random selection supports generalization to the population. Random assignment supports causal conclusions. An observational study with random selection can generalize but cannot establish causation. An experiment with random assignment can establish causation but can only generalize if the units are representative of a larger group. Statistically significant results in a well-designed experiment are evidence that treatments caused the observed effect.
Every inference procedure in Units 6 through 9 assumes that data were collected with randomness. Without random sampling, you cannot generalize to a population. Without random assignment, you cannot claim causation. Bias in sampling or experimental design undermines conclusions regardless of sample size. Understanding why a particular design is or is not appropriate is the core skill of Unit 3 and a recurring task on the AP exam.
Non-random data collection methods such as convenience sampling and voluntary response sampling introduce bias and produce conclusions that cannot be trusted. This topic establishes why randomness is essential for valid statistical conclusions.
Covers the distinction between observational studies and experiments, the relationship between population and sample, and the rules for when generalization and causal conclusions are appropriate.
Introduces the four main random sampling methods: SRS, stratified, cluster, and systematic. Each method has specific mechanics, advantages, and limitations depending on the population structure and research question.
Covers voluntary response bias, undercoverage bias, nonresponse bias, response bias, and measurement bias. The key skill is identifying the specific type of bias in a described sampling situation and explaining its effect on conclusions.
Defines the components of an experiment: experimental units, explanatory variable, response variable, and confounding variable. Describes the four features of a well-designed experiment and introduces completely randomized, block, and matched pairs designs.
Focuses on choosing among completely randomized, randomized block, and matched pairs designs based on the research question, available resources, and the nature of the experimental units. Requires explaining why a design is or is not appropriate.
Connects experimental design to statistical conclusions. Random assignment supports causal claims when results are statistically significant. Random selection of units supports generalization. Both are needed for the strongest conclusions.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
Open this guide for a closer review of the topic.
This snapshot uses Fiveable practice activity to show where students tend to miss questions and which review moves are worth prioritizing first.
Across 9.8k multiple-choice practice attempts for this unit.
Practice activity included in this snapshot.
Across 85 scored free-response attempts for this unit.
Review Selecting an Experimental Design with attention to how the concept appears in AP-style source and evidence questions.
Review Introduction to Experimental Design with attention to how the concept appears in AP-style source and evidence questions.
Review Random Sampling and Data Collection with attention to how the concept appears in AP-style source and evidence questions.
Review Potential Problems with Sampling with attention to how the concept appears in AP-style source and evidence questions.
Data collected without randomness produce conclusions that cannot be trusted. Non-random methods such as convenience sampling and voluntary response sampling introduce systematic bias because the sample is not representative of the population. Recognizing these flaws is the first step in evaluating any study.
| Method | How selected | Main bias risk |
|---|---|---|
| Convenience sample | Whoever is easiest to reach | Undercoverage of hard-to-reach groups |
| Voluntary response sample | Individuals choose to participate | Overrepresentation of strong opinions |
| Simple random sample | Chance selects every group equally | Minimal systematic bias |
The two main study types are observational studies and experiments. In an observational study, researchers record data without imposing treatments. In an experiment, researchers assign treatments to experimental units. The type of study determines whether you can generalize to a population or claim a causal relationship.
| Study type | Treatments imposed? | Can generalize? | Can claim causation? |
|---|---|---|---|
| Observational study | No | Yes, if random sample | No |
| Experiment | Yes | Yes, if units are representative | Yes, if random assignment |
Random sampling uses chance to select individuals, which is what makes generalization to a population valid. The four main methods differ in how the population is structured and how units are selected. Knowing when each method is appropriate and what its trade-offs are is a key exam skill.
| Method | How population is divided | Selection process | Key advantage |
|---|---|---|---|
| SRS | Not divided | Random from whole population | Simple; every group equally likely |
| Stratified | Homogeneous strata | SRS within each stratum | Ensures representation of each group |
| Cluster | Heterogeneous clusters | Random selection of whole clusters | Practical when list is unavailable |
| Systematic | Not divided | Random start, every kth unit | Easy to implement on ordered lists |
Even when a sampling method is chosen carefully, bias can enter through how the sample is assembled or how data are collected. Identifying the specific type of bias and explaining its direction is a common exam task.
| Bias type | Source | Example |
|---|---|---|
| Voluntary response | Self-selection into sample | Online poll where anyone can click to respond |
| Nonresponse | Selected individuals do not respond | Mail survey with 20% return rate |
| Undercoverage | Part of population excluded | Phone survey missing households without landlines |
| Response | Question wording or social pressure | Leading question about a sensitive behavior |
| Measurement | Instrument or method error | Scale that consistently reads 2 lbs too high |
A well-designed experiment has four key features: comparison of at least two treatment groups, random assignment of treatments to experimental units, replication across multiple units per treatment, and control of confounding variables. These features together allow causal conclusions.
| Design feature | Purpose |
|---|---|
| Comparison | Establishes a baseline to measure treatment effects against |
| Random assignment | Balances confounding variables across treatment groups |
| Replication | Reduces the effect of chance variation on results |
| Control of confounding | Prevents extraneous variables from distorting the treatment effect |
The three main experimental designs differ in how they handle variability among experimental units. Choosing the right design depends on the research question, available resources, and whether a meaningful blocking variable exists.
| Design | Blocking used? | Best when | Trade-off |
|---|---|---|---|
| Completely randomized | No | Units are homogeneous | Confounding from unit variability possible |
| Randomized block | Yes, by group | A known variable affects the response | Requires identifying a valid blocking variable |
| Matched pairs | Yes, by pair or within unit | Units can be paired or receive both treatments | Carryover effects possible in within-unit designs |
A well-designed experiment with random assignment allows a causal conclusion when results are statistically significant. Statistical significance means the observed difference is large enough that it is unlikely to have occurred by chance alone. Generalization to a larger population requires that the experimental units be representative of that population.
| Design feature present | Conclusion supported |
|---|---|
| Random assignment only | Causation for the units studied; limited generalization |
| Random selection only | Generalization to population; no causation |
| Both random assignment and random selection | Causation and generalization to population |
| Neither | No causation; no generalization |
Try AP-style multiple-choice questions and written prompts after you review the notes.
| Term | Definition |
|---|---|
| Bias | A systematic error in data collection that causes certain responses to be favored over others, making the sample unrepresentative of the population. |
| Voluntary Response Bias | Bias that occurs when individuals self-select into a sample, typically overrepresenting those with strong opinions and underrepresenting the broader population. |
| Nonresponse Bias | Bias that occurs when individuals selected for a sample do not respond and differ systematically from those who do, distorting the sample results. |
| Observational Study | A study in which researchers observe and record data without imposing treatments; can identify associations but cannot establish causation. |
| Experiment | A study in which researchers assign treatments to experimental units and measure a response variable, allowing causal conclusions when well designed. |
| Stratified Sampling | A sampling method that divides the population into homogeneous strata and takes an SRS within each stratum to ensure representation of each group. |
| Cluster Sample | A sampling method that divides the population into heterogeneous clusters and randomly selects whole clusters to include in the sample. |
| Confounding Variables | A variable related to the explanatory variable that also influences the response variable, potentially creating a false appearance of causation. |
| Random Assignment | The process of allocating treatments to experimental units by chance, which balances confounding variables and supports causal conclusions. |
| Completely Randomized Design | An experimental design in which all experimental units are randomly assigned to treatments without any blocking or grouping. |
| Randomized Block Design | An experimental design that groups units into blocks based on a shared characteristic before randomly assigning treatments within each block to reduce variability. |
| Statistically Significant | A result is statistically significant when it is large enough to be unlikely to have occurred by chance alone, supporting the conclusion that the treatment caused the effect. |
| Statistical Inference | The process of using data from a sample or experiment to draw conclusions about a population or the effect of a treatment. |
| causal conclusion | A conclusion that one variable directly caused a change in another; valid only when random assignment was used in a well-designed experiment. |
In stratified sampling, you take an SRS from every stratum. In cluster sampling, you randomly select whole clusters and include all members. Students often reverse these: strata are homogeneous groups sampled within; clusters are heterogeneous groups selected entirely.
Even a well-designed observational study with a large random sample cannot establish causation. Confounding variables cannot be ruled out without random assignment. The phrase 'causes' or 'leads to' is only justified in a randomized experiment.
Random selection determines who is in the study and supports generalization to the population. Random assignment determines which treatment each unit receives and supports causal conclusions. These are separate design features with separate implications.
Nonresponse bias requires that the non-respondents differ systematically from respondents. A low response rate alone does not guarantee bias; the issue is whether those who did not respond have different characteristics relevant to the study question.
Stratification is a sampling technique used to select a representative sample. Blocking is an experimental design technique used to reduce variability by grouping similar units before random assignment. They use similar logic but apply to different contexts.
A common task presents a study description and asks you to identify whether it is observational or experimental, name the sampling or design method used, and explain whether the design supports generalization, causation, or neither. Precise vocabulary matters: say 'random assignment' not just 'random' when discussing experiments, and name the specific bias type rather than just saying 'biased.'
Free response questions frequently ask you to justify a design choice or critique a flawed one. For sampling, this means explaining why a particular method does or does not produce a representative sample. For experiments, this means explaining how blocking reduces variability or why random assignment controls for confounding. Answers must connect the design feature to its statistical purpose.
A recurring task gives you the results of an experiment and asks what conclusions are supported. You must separately address causation (requires random assignment) and generalization (requires representative units or random selection). Overstating conclusions, such as claiming causation from an observational study or generalizing beyond the sampled population, is a common error that costs points.
Open the individual guides for Unit 3 when you want a closer review of one topic.
browse guidesPractice free-response reasoning and compare your answer with scoring guidance.
practice FRQsWatch past review streams filtered to Unit 3 when you want a video walkthrough.
open videosUse unit cheatsheets for a quick visual review after you work through the notes.
open cheatsheetsEstimate your broader AP score goal after you review the course and exam format.
open calculatorAP Stats Unit 3 covers 7 topics focused on collecting data and experimental design: 3.1 Do the Data We Collected Tell the Truth, 3.2 Introduction to Planning a Study, 3.3 Random Sampling and Data Collection, 3.4 Potential Problems with Sampling, 3.5 Introduction to Experimental Design, 3.6 Selecting an Experimental Design, and 3.7 Inference and Experiments. The big ideas are how to design studies that produce trustworthy data and how to tell the difference between observational studies and experiments. See AP Stats Unit 3 for practice on all seven topics.
AP Stats Unit 3 makes up 12-15% of the AP exam, making it one of the more heavily tested units. It covers collecting data through experimental design and random sampling, including how to identify bias, choose a sampling method, and draw valid conclusions from well-designed studies.
The AP Stats Unit 3 progress check includes both MCQ and FRQ parts drawn from all seven unit topics, with heavy emphasis on experimental design, random sampling methods, and identifying sources of bias. MCQ questions test whether you can recognize study types and spot flaws in data collection. FRQ questions typically ask you to design a study or explain why a method does or does not support a causal conclusion. Topics like 3.4 Potential Problems with Sampling and 3.6 Selecting an Experimental Design show up most often. Practice progress check-style questions at AP Stats Unit 3.
AP Stats Unit 3 FRQs most often ask you to design an experiment or evaluate a sampling method, drawing on topics like 3.5 Introduction to Experimental Design, 3.6 Selecting an Experimental Design, and 3.7 Inference and Experiments. A typical question gives you a scenario and asks you to describe a completely randomized design or a block design, explain how random sampling reduces bias, or state whether a causal conclusion is justified. To practice, write out full responses and check that you name the treatment groups, explain randomization, and address potential confounding variables. You can find FRQ practice aligned to these topics at AP Stats Unit 3.
For AP Stats Unit 3 practice questions, including MCQ and practice test sets, head to AP Stats Unit 3. You'll find multiple-choice questions covering experimental design, random sampling, and bias, plus free-response practice across all 7 topics in the unit. When you work through MCQs, focus on questions that ask you to identify study types and spot problems with data collection methods, since those show up most on the actual exam.
Start AP Stats Unit 3 by building a clear mental map of the difference between observational studies and experiments, since that distinction drives most of the unit. From there, work through these steps: 1. Learn the sampling methods in 3.3 (simple random, stratified, cluster, systematic) and practice explaining why random sampling reduces bias. 2. Study 3.4 Potential Problems with Sampling so you can name and explain undercoverage, nonresponse, and response bias. 3. Work through 3.5 and 3.6 to understand completely randomized designs, block designs, and matched pairs, then sketch out each design type by hand. 4. Finish with 3.7 Inference and Experiments to understand when you can and cannot claim causation. For each topic, write out at least one FRQ-style explanation in your own words. Experimental design questions reward precise vocabulary, so practice using terms like control group, random assignment, and confounding variable correctly. Find practice sets at AP Stats Unit 3.