Confounding Variable Practice for AP Statistics
In this fictional print workshop, 72% of people who chose a quick reference card finished a practice print without help. Among those who chose a full handout, 48% finished on their own. Before you recommend the card, check who chose it.
A confounding variable offers another explanation for an observed association. In an observational study, look for a variable that’s related to both the explanatory variable (what you’re comparing) and the response variable (the outcome you measure). Then ask how that third variable might shift the comparison. College Board’s AP Statistics framework, section 1.10.D.5.

Why the quick card leads overall
Before anyone chose a format, staff recorded who had used a printing press before. Everyone attempted the same print. The explanatory variable is the chosen reference format, and the response is whether the person finished without help.
| Prior print experience | Quick card, finished / total | Full handout, finished / total |
|---|---|---|
| Used a press before | 32/40 (80%) | 8/10 (80%) |
| First time | 4/10 (40%) | 16/40 (40%) |
| All participants | 36/50 (72%) | 24/50 (48%) |
Experienced printers chose the card more often. Of the 50 people who had used a press before, 40 chose the card. Of the 50 first-timers, only 10 did. A quick reminder may be enough when you already know the steps.
Experienced printers were also more likely to finish unaided. Within either format, 80% of experienced printers finished, compared with 40% of first-timers. Anyone who’s used a press before may already know how to line up the paper and apply even pressure.
The groups had identical completion rates at each experience level. The entire 24-percentage-point gap in the observed totals comes from the different mix of participants. The card group was 80% experienced, and the handout group was 20% experienced. Multiply the difference in those shares (0.60) by the 40-percentage-point completion advantage for experienced printers to get 24 percentage points.
Those rates come from one workshop where people chose their own format. Even within the experienced group, people who picked the card may have spent more time at a press than people who picked the handout. Equal rates within these two categories still leave the causal effect uncertain. A conclusion can stay close to what staff recorded.
More quick-card users finished unaided overall. At each experience level, the two formats had the same completion rate. Because participants chose their format, the study leaves open whether switching formats would change how often people finish unaided.
How to write the two links
Name one variable you could actually record, such as whether someone has used a printing press before. Then explain which explanatory group would have more of that variable and why. In a second sentence, explain whether the candidate would tend to raise or lower the response and why. Put the two links together to see which way the comparison could shift. Use could or may for links the scenario leaves unmeasured. Penn State’s introductory statistics lesson, section 1.4.1.
Use measured group counts when you have them. A strong predictor of the response can be evenly distributed across the groups. In that case, the groups’ mix on that measure supplies no explanation for the gap. A detail about how someone joined a group also needs a plausible link to the outcome.
Six practice scenarios
For each scenario, identify the explanatory and response variables. Then propose or assess a candidate confounder, explain the direction of each link where you can, and write a conclusion the design supports.
1. Covered bird feeders
Researchers record more bird visits per hour at covered feeders than at open feeders. Residents chose the feeders for their own gardens. Suggest one garden feature that could affect which feeder people buy and how often birds visit. Explain how it could help produce the observed result.
2. Prior experience in both groups
A craft club compares success assembling a paper model using diagrams or written instructions. Participants choose their format. Prior model-building experience predicts success. Each group has 20 people, exactly 8 with prior experience and 12 without. The diagram group has a higher success rate. Could a difference in the groups’ experience mix explain the gap?
3. A color on the order form
A bakery compares how often shoppers buy a loaf after choosing a sample cut into slices or left in one piece. Staff tend to use blue forms for shoppers who choose sliced samples because those forms are stacked next to the slicer. They fill in and file the forms out of the shopper’s sight after the shopper decides whether to buy. Could the color explain a difference in purchase rates?
4. Sampled at random, route chosen by riders
A city takes a random sample from its complete bike-share membership list. Everyone selected takes one ride between the same two stations and chooses either a park route or a street route. Park-route riders record shorter trips on average. Experienced cyclists tend to choose the park route and may pedal faster. How could experience affect the comparison, and what does the random sampling help with?
5. Feedback and extra time
A teacher randomly assigns students to two study sessions. One group gets written feedback and 30 minutes to practice. The other gets no feedback and 15 minutes. Both take the same quiz, and the 30-minute group scores higher on average. Which treatment comparison did the teacher randomize? How could the teacher isolate the effect of feedback at a fixed practice time?
6. Mesh covers and insect damage
Gardeners tend to put mesh over plants already facing heavy insect pressure. Heavier initial pressure tends to lead to more damaged leaves by the end of the week. In the observed data, covered plants have fewer damaged leaves. Could initial insect pressure confound this comparison? Which way would it tend to push the covered group’s damage counts?
Answers and explanations
-
The explanatory variable is feeder type, and the response is visits per hour. One possible confounder is nearby shrubs. Gardeners may prefer covered feeders under shrubs that drop twigs. Shrubs can also give birds places to perch or hide, encouraging more visits. This could make visits more frequent at covered feeders even if the covers themselves have no effect. The study shows an association between feeder type and visit rate in these gardens.
-
The explanatory variable is instruction format, and the response is whether the participant assembles the model successfully. Both groups are 40% experienced and 60% beginners. Prior experience predicts success, and its measured distribution is the same across formats. The overall gap comes from differences in observed success within one or both experience groups. The gap might be larger for beginners, for example, which you could check by comparing the formats separately at each experience level. People chose their format, so other differences among participants could still affect those comparisons.
-
The explanatory variable is sample format, and the response is whether the shopper buys a loaf. Order-form color tracks the sample format because of where staff keep the forms. It’s a clerical detail recorded out of sight after the purchase decision, so it supplies no plausible cause of that decision. The study can describe the association between sample format and purchases at this bakery.
-
The explanatory variable is route choice, and the response is trip time. Prior cycling experience is more common among riders who choose the park route and could shorten trip times on either route. This could explain part or all of the observed difference. The random sample gives a basis for generalizing the association to the city’s bike-share members making this trip under the study conditions. Riders chose their route, which leaves the route’s causal effect uncertain. More examples on random sampling and random assignment.
-
The response is quiz score. The teacher randomized two combinations of feedback and practice time. Since feedback always came with 30 minutes of practice, a higher score could come from the feedback, the extra time, or how they worked together. With sound conduct and an analysis that accounts for chance variation, the experiment can support a causal comparison of the two sessions as delivered. To test feedback at a fixed duration, give everyone 30 minutes and randomly assign who receives feedback. College Board’s framework on experimental design, sections 1.13.A.7–10.
-
The explanatory variable is mesh use, and the response is the number of damaged leaves at the end of the week. Initial insect pressure is a plausible confounder. Plants under heavier pressure were more likely to get mesh, and heavier pressure tends to increase damage. That would tend to push the covered group’s damage counts up, in the opposite direction to the observed gap. Covered plants had fewer damaged leaves in this study, and the effect of the mesh itself remains uncertain.
Two recall cards
| Front | Back |
|---|---|
| What two relationships should you explain when you propose a confounder? | How the candidate relates to the explanatory variable and how it could affect the response. Give the direction of each link in the context of the study. |
| A variable predicts success, and its distribution is the same in both groups. What does that tell you about the overall success gap? | Both groups have the same mix on that measure, so an imbalance in it can’t explain the gap. |
Cover the answers and recall them from memory today, tomorrow, and again a few days later. Spacing out that recall gives you practice retrieving the two links after some time has passed. Each time, redo one of the scenarios in full sentences. For broader course review, the deck also includes prompts on confounding. English AP Statistics flashcard deck.
AI assisted with drafting the text and exercises and generated the cover image.