Skip to content
AP Statistics · Exploring One-Variable Data and Collecting Data

Confounding

Confounding appears in 90 reviewed AP Statistics questions and sits in Exploring One-Variable Data and Collecting Data. Below are worked examples showing exactly how the exam tests it, with the answer marked and every distractor explained.

How Confounding is tested

Example 1Harder
In an experiment testing a new cognitive training program, participants are randomly assigned to either the training group or a control group. The control group spends the same amount of time on unrelated puzzles. Why is having the control group do puzzles (rather than nothing) important?
  1. AIt ensures all participants benefit from the study.
  2. BIt creates a matched pairs design with the training group.
  3. CIt controls for the Hawthorne effect.Correct
  4. DIt reduces the sample size needed for significance.

Why C is right

The Hawthorne effect refers to participants changing behavior or performance simply because they are being observed or participating in a study, regardless of the specific treatment. Participants may improve simply from receiving attention and doing activities, not from the specific training. By having the control group also engage in an activity (unrelated puzzles) for the same duration, the researcher controls for this effect — any improvement in the training group above the control group can be attributed to the specific training, not just the attention or engagement.
Why the other 3 choices miss
A
Having control participants do puzzles is not about ensuring they benefit. The purpose is methodological: to provide a valid comparison for the training group.
B
A matched pairs design pairs individual participants and assigns different treatments within each pair. Using a separate control group with a comparable activity is not a matched pairs design.
D
The type of control activity does not affect the sample size needed for statistical significance. Sample size is determined by the desired power and effect size.
Example 2Medium
A researcher wants to determine if a new app reduces screen time. She allows users to choose whether to download the app (treatment group) or not (control group). After one month, she compares screen time between the two groups. The primary reason this is not a valid experiment is:
  1. AThe control group should have received a fake app.
  2. BThe groups were self-selected, not randomly assigned, so pre-existing differences (confounding) prevent causal conclusions.Correct
  3. CScreen time cannot be measured accurately.
  4. DThe study is too short at one month.

Why B is right

Without random assignment, this is an observational study, not an experiment. Users who choose to download the app may already be more motivated to reduce screen time. This self-selection creates confounding — the observed reduction in screen time could be due to user motivation, not the app. Random assignment would have controlled for this.
Why the other 3 choices miss
A
A fake app (active placebo) would be an improvement, but the most fundamental flaw is the lack of random assignment, which allows self-selection bias and confounding.
C
Screen time can be measured accurately using phone usage statistics. Measurement accuracy is not the primary concern described.
D
Duration is a study design choice but is not the fundamental flaw here. Even a one-month experiment can produce valid causal evidence if random assignment is used.
Example 3Medium
A research team studies the effect of tutoring on math achievement. They compare students who voluntarily attended tutoring sessions with students who did not. The tutored students scored higher on the final exam. The researchers report that tutoring causes improved math scores. A statistician identifies a likely confounding variable. Which of the following is the most plausible confounding variable?
  1. AThe time of day when the exam was administered
  2. BThe color of the test paper used for the exam
  3. CStudents' motivation to learn mathCorrect
  4. DThe number of exam questions

Why C is right

Student motivation is a plausible confounding variable because it is associated with both the explanatory variable (seeking tutoring) and the response variable (test scores). Motivated students are more likely to attend tutoring AND more likely to study harder and score well, regardless of tutoring. Without random assignment, motivation confounds the comparison.
Why the other 3 choices miss
A
Incorrect because the time of the exam is presumably the same for all students and would not differentially affect tutored versus non-tutored students.
B
Incorrect because test paper color is unlikely to affect performance and is presumably the same for all students.
D
Incorrect because the number of exam questions is the same for all students and does not vary between groups.

Studied alongside

Other concepts from Exploring One-Variable Data and Collecting Data.