K-12 Math (US): 9-12 Data Analysis K-12 Math Statistics Interpreting scatterplots

Study Guide: Interpreting Scatterplots (Grade 9–12, Math)

1. The Driving Question

"If you plot every student’s shoe size against their test scores, why do the dots sometimes form a line, sometimes a cloud, and sometimes nothing at all—and how can you tell if one thing is actually causing the other?" This isn’t just about drawing dots on a graph; it’s about reading the hidden stories in messy real-world data. By the end, you’ll know how to spot patterns, measure their strength, and avoid jumping to conclusions like "bigger shoes = smarter kids."

2. The Core Idea — Built, Not Listed

Imagine you’re at a county fair, and a carnival game claims that the longer you hold your breath, the higher you’ll score on a ring-toss game. To test this, you record 20 players’ breath-holding times (in seconds) and their ring-toss scores. You plot each player as a dot on a graph: x-axis = breath-holding time, y-axis = score. Now, the dots don’t form a perfect line, but they lean upward—most players who held their breath longer scored higher. That lean is the direction of the relationship. How tightly the dots hug an imaginary line is the strength—if they’re scattered like confetti, the relationship is weak; if they’re packed like subway riders at rush hour, it’s strong. But here’s the catch: even if the dots form a perfect line, that doesn’t mean holding your breath causes better scores. Maybe players who practice more have better lung capacity and better aim. The scatterplot shows association, not causation—a distinction that trips up even professional researchers.

Key Vocabulary:
- Scatterplot: A graph where each dot represents one observation, with the x-axis showing one variable and the y-axis showing another.
Example: Plotting the number of hours a YouTuber streams per week (x) against their subscriber count (y).
College shift: In advanced statistics, scatterplots are used to visualize multivariate relationships (e.g., 3D plots or color-coding a third variable).

• Correlation (r): A number between –1 and 1 that measures the direction (positive/negative) and strength (how close to –1 or 1) of a linear relationship between two variables.
  Example: The correlation between ice cream sales and drowning incidents is strong and positive—not because ice cream causes drowning, but because both increase in hot weather.
  College shift: Correlation assumes linearity; in college, you’ll learn to test for nonlinear relationships (e.g., U-shaped or exponential).

• Outlier: A data point that falls far outside the pattern of the other dots, often skewing results or hinting at an error (or a fascinating exception).
  Example: In a scatterplot of NBA players’ heights vs. points per game, a 5’9" player averaging 30 points would be an outlier (e.g., Isaiah Thomas).
  College shift: Outliers are analyzed using robust statistics (e.g., median absolute deviation) to reduce their influence on models.

• Lurking variable: A hidden third variable that influences both the x and y variables, creating a misleading association.
  Example: A scatterplot showing a positive correlation between shoe size and reading ability in children might suggest bigger feet = better readers—but age is the lurking variable (older kids have bigger feet and better reading skills).
  College shift: In causal inference, lurking variables are controlled for using techniques like randomized experiments or regression adjustment.

3. Assessment Translation

How this appears on assessments:
- SAT/ACT: Multiple-choice questions testing interpretation of scatterplots (e.g., "Which scatterplot shows the strongest negative correlation?") or calculating correlation coefficients. Distractors often include: - Confusing direction (positive vs. negative) with strength (strong vs. weak).
- Ignoring outliers that distort the pattern.
- Misinterpreting causation (e.g., "Does this prove x causes y?").
- AP Statistics: Free-response questions (FRQs) where you must: 1. Describe the form (linear/nonlinear), direction, and strength of a relationship.
2. Identify outliers and explain their potential impact.
3. Interpret the slope of a regression line in context (e.g., "For each additional hour of study, test scores increase by 5 points").
Rubric priorities: Contextual explanations (not just "r = 0.8") and justifying why correlation ≠ causation.

Proficient vs. Developing Responses:
| Prompt | Developing Response | Proficient Response | |------------|-------------------------|-------------------------| | "Describe the relationship between hours of sleep and test scores in the scatterplot below." | "The dots go up, so more sleep means better scores." | "The scatterplot shows a moderate, positive, linear relationship: as hours of sleep increase, test scores tend to increase. However, there’s one outlier—a student who slept 9 hours but scored poorly, which might indicate another factor (e.g., illness) affected their performance. The correlation isn’t perfect, so sleep alone doesn’t guarantee higher scores." |

Model Proficient Response (AP FRQ):
Prompt: A researcher plots the number of hours high school students spend on social media (x) against their GPA (y). The scatterplot shows a weak negative correlation (r = –0.3). The researcher concludes, "Social media harms academic performance." Do you agree? Justify your answer.

Response: The researcher’s conclusion is too strong. While the scatterplot shows a weak negative correlation (r = –0.3), this only suggests a slight tendency for students who use social media more to have lower GPAs—not that social media causes lower grades. A lurking variable, like time management skills, could explain the pattern: students who struggle with time management might spend more time on social media and have lower GPAs. To test causation, the researcher would need a controlled experiment (e.g., randomly assigning students to limit social media use). Additionally, the weak correlation means other factors (e.g., study habits, sleep) likely play a larger role in GPA.

4. Mistake Taxonomy

Mistake 1: Confusing Correlation with Causation
Prompt: A scatterplot shows a strong positive correlation between ice cream sales and shark attacks. A student writes, "Eating ice cream causes shark attacks." - Wrong response: "The dots form a line, so ice cream must attract sharks." - Why it loses credit: The response ignores lurking variables (e.g., hot weather increases both ice cream sales and beach attendance, where shark attacks are more likely). Causation requires evidence beyond correlation.
- Correct approach: 1. Acknowledge the correlation: "There is a strong positive association between ice cream sales and shark attacks." 2. Identify a lurking variable: "Hot weather likely increases both ice cream sales and the number of people swimming in the ocean." 3. Reject causation: "This does not prove ice cream causes shark attacks; it’s a coincidence driven by a third factor."

Mistake 2: Overlooking Outliers
Prompt: A scatterplot shows a strong positive correlation (r = 0.9) between study hours and test scores, but one student studied 10 hours and scored 50%. The student writes, "The correlation is strong, so study time determines test scores." - Wrong response: "The correlation is 0.9, so the relationship is almost perfect." - Why it loses credit: The outlier (10 hours, 50% score) suggests the relationship isn’t consistent for all students. Ignoring it overstates the strength of the pattern.
- Correct approach: 1. Note the outlier: "One student studied 10 hours but scored 50%, which doesn’t fit the pattern." 2. Explain its impact: "This outlier may artificially inflate the correlation. Without it, the relationship might be weaker." 3. Contextualize: "The outlier could represent a student with a learning disability or test anxiety, showing that study time alone doesn’t guarantee high scores."

Mistake 3: Misinterpreting "No Correlation"
Prompt: A scatterplot of shoe size vs. favorite color shows no pattern. A student writes, "There is no relationship between shoe size and favorite color." - Wrong response: "The correlation is 0, so shoe size and favorite color are unrelated." - Why it loses credit: "No correlation" only means no linear relationship. The student fails to consider nonlinear patterns (e.g., favorite colors might cluster by shoe size in a U-shape).
- Correct approach: 1. Clarify the type of relationship: "The scatterplot shows no linear correlation (r ≈ 0)." 2. Acknowledge other possibilities: "However, there could be a nonlinear relationship (e.g., people with very small or very large shoes might prefer the same color)." 3. Conclude cautiously: "Without further analysis, we can only say there’s no linear association."

5. Connection Layer

1. Within Math: Scatterplots → Regression lines
   Why it matters: A scatterplot shows the pattern; a regression line quantifies it (e.g., "For each additional hour of sleep, GPA increases by 0.1 points"). Understanding scatterplots makes regression intuitive—you’re just drawing the "best-fit" line through the dots.

2. Across Subjects: Scatterplots → Epidemiology (Science)
   Why it matters: In public health, scatterplots track relationships like smoking rates vs. lung cancer cases. The same logic applies: a strong correlation prompts further study (e.g., "Does smoking cause cancer?"), but lurking variables (e.g., genetics, pollution) must be ruled out.

3. Outside School: Scatterplots → Sports Analytics
   Why it matters: NBA teams use scatterplots to analyze player performance (e.g., minutes played vs. points scored). Outliers (e.g., a bench player with high efficiency) can reveal undervalued talent—or flaws in the data (e.g., small sample size). Next time you see a "breakout star," ask: Is this a real pattern or just noise?

6. The Stretch Question

"If you plot the number of pirates in the world against global average temperature, you’ll find a strong negative correlation. Does this mean pirates cool the planet? How would you design a study to test this (ridiculous) claim?"

Pointer toward the answer: This is a classic example of a spurious correlation—a relationship that appears meaningful but is actually coincidental. To test it, you’d need to: 1. Rule out lurking variables: Identify factors that might influence both piracy and temperature (e.g., industrialization reduced piracy and increased CO₂ emissions).
2. Control for time: Use time-series analysis to see if changes in piracy precede changes in temperature (a requirement for causation).
3. Propose a mechanism: If pirates did cool the planet, how? (e.g., "Pirate ships block sunlight" is absurd, but it’s the kind of testable hypothesis you’d need.) The takeaway: Correlation is a starting point, not proof. The fun part of statistics is asking, "What else could explain this?"—even when the claim is silly.