Association in AP Statistics

In AP Statistics, two variables have an association when knowing the value of one changes what you'd expect for the other. For categorical variables, the conditional distributions differ across groups. For quantitative variables, the scatterplot shows a pattern with a direction, form, and strength.

What is Association?

Two variables are associated if learning one tells you something about the other. That's the whole idea. If knowing someone's education level shifts the odds of their voting preference, those two variables are associated. If it doesn't change anything, they aren't, and the variables are independent.

How you check for association depends on the variable type. With two categorical variables, you compare conditional relative frequencies (percentages within each group). If the percentages look about the same across groups, there's no association. If they differ noticeably, there is. You can see this in a contingency table, a segmented bar graph, or a mosaic plot. With two quantitative variables, you look at a scatterplot and describe the pattern by its direction, form, strength, and any unusual features. Either way, association describes a relationship. It never says one variable causes the other.

Why Association matters in AP® Statistics

Association is the main question behind all of two-variable data in AP Stats. In Topics 2.1 and 2.2, you build and read two-way tables and graphs, and the point is to decide whether two categorical variables are associated. In Topic 5.1, the same question moves to scatterplots of two quantitative variables. Topic 3.14 then turns the eyeball judgment into formal inference. A chi-square test for independence uses "no association" as its null hypothesis. So association shows up in exploring data, in inference, and in study design (association vs. causation). Released FRQs have hit all of these.

How Association connects across the course

Contingency Tables and Conditional Distributions (Unit 2)

A contingency table is the raw evidence, and conditional percentages are how you read it. If 60% of SUV crashes are severe but only 30% of sedan crashes are, the conditional distributions differ, and that difference is the association. Always compare percentages, not raw counts, because the groups may be different sizes.

Chi-Square Test for Independence (Unit 3)

The chi-square test is association put on trial. The null hypothesis says there's no association (the variables are independent), and the alternative says there is one. Expected counts show what the table would look like with zero association. A big gap between observed and expected counts is evidence of a real relationship, not just sampling variability.

Scatterplots and Two Quantitative Variables (Unit 5)

The same idea carries over to numbers. Instead of comparing percentages, you ask whether y tends to rise or fall as x increases. A positive, linear, strong pattern (like bullfrog length vs. mass) is just association described in scatterplot language.

Simpson's Paradox (Unit 2)

An association can flip direction once you split the data by a third variable. Simpson's Paradox is the warning label on every association you find. What looks true for the combined data might reverse inside each subgroup because a lurking variable was hiding the real story.

Is Association on the AP® Statistics exam?

Association gets tested in three main ways.

Reading displays (MCQ). Expect stems built around mosaic plots, segmented bar graphs, and two-way tables. You might be asked what it means when a mosaic plot cell (like SUVs with severe crashes) is larger than it would be under independence. That signals association. You might also be asked what unequal column widths mean. Those just reflect different group sizes (the marginal distribution of one variable), not association. Another common stem compares segmented bar graphs to mosaic plots. Both show conditional distributions, but only mosaic plots also show group sizes through their widths.

Describing and testing (FRQ). Released FRQs such as 2017 Q5 (age at diagnosis for men and women) ask whether data provide convincing evidence of a difference or relationship, which calls for a full inference procedure with hypotheses in context. FRQs like 2022 Q1 (bullfrog length and mass) and 2023 Q5 (tule elk weight) involve relationships between quantitative variables, where you describe the association in a scatterplot and use it for prediction.

Study design. FRQs like 2021 Q2 (walking and cholesterol) test whether you can tell when a study supports only an association and when it supports a cause-and-effect claim. Observational data give you association. Only a randomized experiment supports causation.

Association vs Causation

Association means two variables move together. Causation means changing one actually produces a change in the other. An observational study can show association, but a lurking variable might explain it. Only a well-designed randomized experiment lets you conclude causation. On the exam, saying "X causes Y" based on observational data or a correlation will cost you points.

Key things to remember about Association

• Two variables are associated when knowing the value of one changes what you'd expect for the other.

• For categorical variables, compare conditional relative frequencies, and if they differ noticeably across groups, the variables are associated.

• For quantitative variables, describe the association in a scatterplot using direction, form, strength, and unusual features, all in context.

• A chi-square test for independence uses "no association between the variables" as its null hypothesis.

• Unequal column widths in a mosaic plot show different group sizes, not an association between the variables.

• Association is not causation, and only a randomized experiment can support a cause-and-effect conclusion.

Frequently asked questions about Association

What is association in AP Statistics?

Association means two variables are related, so knowing one gives you information about the other. For categorical data, it shows up as different conditional distributions across groups. For quantitative data, it shows up as a pattern in a scatterplot.

Does association mean one variable causes the other?

No. Association only says the variables move together, and a lurking variable could be behind the pattern. You need a randomized experiment, like the kind of design asked about in 2021 FRQ Q2, to support a causal claim.

How do you tell if two categorical variables are associated?

Calculate conditional relative frequencies (percentages within each group) and compare them. If the percentages are about the same, there's no association. If they differ, there is. To decide whether the difference is statistically convincing, run a chi-square test for independence.

What's the difference between association and independence?

They're opposites. Independent variables have identical conditional distributions across groups, so knowing one tells you nothing about the other. Associated variables don't. In a chi-square test, independence is the null hypothesis and association is the alternative.

Do unequal widths in a mosaic plot mean the variables are associated?

No. The widths only show how many observations fall in each group of the first variable, which is its marginal distribution. Association shows up when the segment heights within each column differ from column to column.