Generalization in AP Statistics

In AP Statistics, generalization is extending results from a sample to a larger population, which is only appropriate when the sample was randomly selected from (or is otherwise representative of) that population, and only to the specific population the sample came from.

What is generalization?

Generalization answers one question. Who do these results actually apply to? When you collect data from a sample, you usually want to say something about a bigger group, the population. The CED is strict about when you're allowed to do that. You can only generalize if the sample was randomly selected from the population, or is otherwise representative of it. And here's the part everyone forgets, you can only generalize to the population the sample was actually drawn from, not some broader group you wish you had sampled.

So if you randomly select 100 adults from one city, your results generalize to adults in that city. Not to all adults, not to all Americans, just that city. If your sample is volunteers, people at a fitness center, or whoever clicked a Twitter link, you can't safely generalize to anyone beyond that self-selected group. Generalization is one half of "scope of inference." The other half is causation, which depends on random assignment instead of random selection. Keep those two jobs separate and Unit 3 gets a lot easier.

Why generalization matters in AP® Statistics

Generalization lives in Topic 3.2 (Introduction to Planning a Study) in Unit 3: Collecting Data, and it directly supports learning objective 3.2.B, identifying appropriate generalizations and determinations based on observational studies. The essential knowledge spells it out in two rules. First, generalizing is only appropriate with random or representative samples. Second, a sample is only generalizable to the population it was selected from. This is also where the CED draws the line on observational studies, since no treatments are imposed, you can never conclude cause and effect from one, no matter how big or random the sample is. Scope-of-inference reasoning is one of the most reliably tested ideas in all of AP Stats, showing up in study-design FRQs almost every year.

How generalization connects across the course

Random Assignment (Unit 3)

These are the two halves of scope of inference. Random selection lets you generalize to a population, while random assignment lets you conclude cause and effect. A study can have one, both, or neither, and the exam loves making you figure out which conclusions are allowed.

Observational Study (Unit 3)

An observational study with a random sample can be generalized to its population, but it can never prove causation because treatments weren't imposed. So observational studies max out at "these variables are associated in this population."

Population (Unit 3)

Generalization is meaningless until you name the population. The sampled population (where the sample actually came from) is the only group your results extend to, even if the researcher's target population is bigger.

Confounding (Unit 3)

Generalization and confounding are separate problems. A random sample fixes generalization but does nothing about confounding variables, which is why a perfectly random survey still can't tell you that one variable causes another.

Randomness condition (Units 6-9)

When you run inference procedures later in the course, the "random" condition you check is generalization showing up again. A confidence interval or significance test is only trustworthy for the population the data was randomly drawn from.

Is generalization on the AP® Statistics exam?

Multiple-choice questions hand you a flawed sampling method and ask which population the results can be generalized to. The pattern is consistent. A researcher samples fitness center members in Chicago, college students at one Texas university, or Twitter users who clicked a survey link, then tries to claim the results apply to "all adults" or "young consumers." The correct answer always shrinks the conclusion back to the group actually sampled. On FRQs, this shows up as a scope-of-inference part inside study-design questions, like the 2021 question about a random sample of 100 adults in a walking and cholesterol study or the 2023 question about fibers in concrete driveways. You're expected to state explicitly whether results generalize and why, naming the random selection (or lack of it) and the specific population. A vague "yes, because it's random" won't earn full credit. Say what kind of randomness it was and which population it justifies.

Generalization vs Causation (cause-and-effect conclusions)

Generalization and causation are answered by different design features, and mixing them up is the classic Unit 3 error. Random selection of the sample controls generalization, meaning who the results apply to. Random assignment of treatments controls causation, meaning whether one variable caused changes in another. An observational study with a beautifully random sample still cannot establish causation, and an experiment with random assignment but volunteer subjects still cannot generalize beyond those volunteers. Check each design feature separately and answer each question separately.

Key things to remember about generalization

• Generalization means extending sample results to a larger population, and it's only appropriate when the sample was randomly selected or otherwise representative of that population.

• A sample is only generalizable to the population it was actually selected from, so a random sample of Chicago fitness center members tells you about Chicago fitness center members, not all adults.

• Random selection justifies generalization; random assignment justifies cause-and-effect conclusions. These are two separate questions answered by two separate design features.

• Observational studies can never establish causal relationships between variables, even when the sample is random and the association is strong.

• Voluntary response samples, like online surveys posted on social media, generalize to almost no one because the respondents selected themselves.

Frequently asked questions about generalization

What is generalization in AP Stats?

Generalization is extending results from a sample to a larger population. Per the CED (Topic 3.2), it's only appropriate when the sample was randomly selected from or is otherwise representative of that population, and results only extend to the population the sample came from.

Does a large sample size let you generalize?

No. Sample size doesn't fix a biased selection method. A voluntary response survey of 10,000 Twitter users is still only generalizable to people like those respondents, while a random sample of 100 adults from a target population generalizes just fine to that population.

What's the difference between generalization and causation?

Generalization is about who the results apply to and requires random selection of the sample. Causation is about whether one variable caused changes in another and requires random assignment of treatments in an experiment. A study can support one, both, or neither.

Can you generalize from an observational study?

Yes, if the sample was randomly selected, you can generalize the observed association to the population sampled. What you can never do with an observational study is conclude that one variable caused the other, since no treatments were imposed.

How do I answer a scope-of-inference FRQ question?

State whether the sample was randomly selected and name the exact population it came from, then conclude that results generalize only to that population. For example, on the 2021 FRQ about walking and cholesterol, the random sample of 100 adults from the target population justified generalizing to that target population specifically.