MATH 2 • STATISTICS & PROBABILITY

Interpreting Correlation — I can interpret correlation and explain limitations of using correlation to make causal claims.

Understanding why two variables moving together doesn't always mean one causes the other.

Historical Context & Motivation

Humans are natural pattern seekers. For centuries, scientists, doctors, and everyday people noticed that certain things seemed to go together — cities with more hospitals also had more disease, countries that ate more chocolate won more Nobel Prizes. But noticing a pattern is not the same as understanding why it exists. The mathematical study of correlation gave us a precise tool for measuring how strongly two variables move together — and an important reminder that patterns don't always mean what we think they mean.

1888
Galton's Insight
Sir Francis Galton published his pioneering study of the relationship between parents' and children's heights, introducing the concept of co-relation — the idea that two measurements could be systematically linked.
1896
Pearson's Formula
Karl Pearson developed the correlation coefficient (r), giving researchers a single number between −1 and 1 to quantify linear relationships between two variables.
1950s
Smoking & Cancer Debate
The tobacco industry famously argued that the strong correlation between smoking and lung cancer was not proof of causation. This public debate made the phrase correlation does not imply causation one of the most important ideas in modern statistics.
2000s
Big Data & Spurious Correlations
With massive datasets, researchers found absurd correlations — such as the divorce rate in Maine correlating with margarine consumption. These examples highlighted the critical need to distinguish correlation from causation.

The central question this lesson addresses is both simple and profound: when two variables appear connected, how do we measure that connection, and how do we avoid the dangerous mistake of assuming one variable causes the other just because they move together?

Core Principles & Definitions

Before diving into calculations and examples, you need a solid grasp of the key ideas that underpin correlation analysis. These four principles form the foundation of everything that follows.

1

Correlation Coefficient (r)

A numerical value between −1 and +1 that measures the strength and direction of a linear relationship between two quantitative variables. Values near ±1 indicate strong linear patterns; values near 0 suggest weak or no linear relationship.
2

Direction of Association

A positive correlation (r > 0) means both variables increase together. A negative correlation (r < 0) means one variable increases as the other decreases.
3

Strength of Association

The absolute value |r| tells you how tightly the data points cluster around a line. Generally, |r| > 0.7 is considered strong, 0.3–0.7 is moderate, and |r| < 0.3 is weak. These thresholds vary by field.
4

Correlation ≠ Causation

Even a very strong correlation between two variables does not prove that one causes the other. A lurking variable, reverse causation, or pure coincidence could explain the pattern.
KEY TAKEAWAY
Think of correlation like a thermometer: it tells you the temperature, but it doesn't tell you why it's hot outside. A strong r-value tells you that two variables move together, but you still need additional evidence — like a controlled experiment — to claim one variable actually causes the other to change.

Visualizing Correlation with Scatterplots

The best way to understand correlation is to see it. A scatterplot shows each data pair as a dot on a coordinate plane. The overall shape of the dot cloud tells you about the direction and strength of the relationship. The diagram below shows four different scatterplot patterns, each with a different value of r.

Four scatterplot panels showing different correlation strengths. Notice how the dots cluster more tightly around the trend line as |r| approaches 1, and scatter widely when r is near 0. The spectrum bar below summarizes the full range of possible r-values from −1 to +1.

In the first panel, the dots hug the upward-sloping dashed line tightly, showing a strong positive correlation. The second panel shows the same tight clustering but sloping downward — a strong negative correlation. The third panel has dots scattered more loosely around the line, representing a weak positive correlation. Finally, the fourth panel shows dots sprayed almost randomly, with virtually no linear correlation. Remember: r only measures linear relationships. Data could follow a perfect curve and still have r ≈ 0.

The Correlation Coefficient Formula

The Pearson correlation coefficient (r) quantifies the linear relationship between two variables x and y. While your calculator or software typically computes this for you, understanding the formula helps you see what r is really doing: comparing how far each data point falls from its mean, and checking whether x and y deviate in the same direction at the same time.

PEARSON CORRELATION COEFFICIENT
r = (1 / (n − 1)) × Σ [(xᵢ − x̄) / sₓ] × [(yᵢ − ȳ) / sᵧ]
Where n = number of data pairs, x̄ and ȳ are the means of x and y, sₓ and sᵧ are the standard deviations of x and y, and the sum runs over all data pairs i = 1 to n.

Let's break this down. The expression (xᵢ − x̄) / sₓ is the z-score of each x-value — it tells you how many standard deviations that value is from the mean. The same applies to y. When you multiply these z-scores together, you get a positive product if both x and y are on the same side of their means (both above or both below), and a negative product if they are on opposite sides.

COEFFICIENT OF DETERMINATION
r² = (fraction of variation in y explained by the linear relationship with x)
For example, if r = 0.80, then r² = 0.64, meaning 64% of the variation in y can be accounted for by the linear relationship with x. The remaining 36% is due to other factors or randomness.
📌 Important Note
The value of r does not change if you switch which variable is x and which is y. It also does not change if you rescale or shift either variable (for example, converting temperature from Fahrenheit to Celsius). Correlation is unit-free.

Why Correlation Does Not Imply Causation

This is the most critical idea in this lesson. When you find a strong correlation between two variables, there are several possible explanations — and direct causation is only one of them. Understanding these alternatives will protect you from making faulty conclusions.

Three possible explanations when variables A and B are correlated: (1) A directly causes B, (2) a hidden third variable C causes both A and B, or (3) the pattern is a random coincidence with no meaningful connection.

A lurking variable (also called a confounding variable) is a hidden factor that influences both of the variables you are studying. In the classic example, ice cream sales and drowning deaths are strongly correlated — but ice cream doesn't cause drowning. Instead, hot weather drives people both to buy ice cream and to go swimming, increasing the risk of drowning. The lurking variable (temperature) creates the illusion of a direct connection between two variables that are not causally linked at all.

There is also the possibility of reverse causation: maybe B causes A, not the other way around. For instance, you might find that people who own fire extinguishers have more fires in their homes. But owning an extinguisher doesn't cause fires — people who have already experienced or fear fires are more likely to buy extinguishers. Finally, with large datasets, you can find striking correlations that are nothing more than spurious coincidences. The only reliable way to establish causation is through a carefully designed controlled experiment with random assignment.

Worked Example: Study Hours and Test Scores

A teacher collects data from 8 students, recording how many hours each student studied for a math test and the score they received. She wants to find and interpret the correlation coefficient.

Study hours and test score data for 8 students
StudentHours Studied (x)Test Score (y)
1152
2258
3365
4470
5574
6679
7783
8890
Finding and Interpreting the Correlation
1
Step 1 — Calculate the MeansFind the mean of x: x̄ = (1 + 2 + 3 + 4 + 5 + 6 + 7 + 8) / 8 = 36 / 8 = 4.5. Find the mean of y: ȳ = (52 + 58 + 65 + 70 + 74 + 79 + 83 + 90) / 8 = 571 / 8 = 71.375.
x̄ = 4.5, ȳ = 71.375
2
Step 2 — Calculate Standard DeviationsUsing the formula for sample standard deviation, we get sₓ ≈ 2.449 and sᵧ ≈ 12.518. These measure how spread out each variable is from its own mean.
sₓ ≈ 2.449, sᵧ ≈ 12.518
3
Step 3 — Compute the z-Score ProductsFor each student, compute the z-score of x, the z-score of y, and multiply them together. For example, Student 1: zₓ = (1 − 4.5) / 2.449 ≈ −1.429, zᵧ = (52 − 71.375) / 12.518 ≈ −1.549, product ≈ 2.213. Repeat this for all 8 students and sum the products. The sum of all z-score products ≈ 6.879.
Σ(zₓ × zᵧ) ≈ 6.879
4
Step 4 — Apply the FormulaDivide by (n − 1): r = 6.879 / (8 − 1) = 6.879 / 7 ≈ 0.983. Using technology (such as a graphing calculator or spreadsheet) confirms this value.
r ≈ 0.983
5
Step 5 — Interpret the ResultAn r-value of 0.983 is very close to 1, indicating a strong positive linear correlation. As study hours increase, test scores tend to increase in a nearly straight-line pattern. However, we must be careful: this data alone does not prove that studying more causes higher scores. Other factors — motivation, prior knowledge, tutoring — could be lurking variables. To make a causal claim, we would need a controlled experiment.
Strong positive linear correlation; causation is not established.

Strengths and Limitations of Correlation

Correlation is an incredibly useful statistical tool, but like any tool, it works best when you know what it can and cannot do. The table below summarizes its strengths and limitations.

Strengths versus limitations of the correlation coefficient
StrengthsLimitations
Provides a single, easy-to-interpret number (r) that summarizes the strength and direction of a linear relationship.Only measures linear relationships. A perfect curved pattern (like a parabola) can yield r ≈ 0.
Unit-free: r does not depend on the scale of measurement, making comparisons across different studies straightforward.Sensitive to outliers. A single extreme data point can dramatically inflate or deflate r.
Easy to compute with technology and widely understood across disciplines.Cannot establish causation. A strong r tells you nothing about whether x causes y, y causes x, or a lurking variable causes both.
r² gives a clear percentage of variation explained, aiding practical interpretation.Requires both variables to be quantitative. It does not apply to categorical data (like gender or favorite color).
KEY TAKEAWAY
Think of correlation like a drone that flies high above a maze. From above, the drone can see the shape of the path (the pattern), but it cannot tell you what is inside the walls or why the path curves the way it does. To figure out the 'why,' you need to come down to ground level and run a controlled experiment.

Connecting to Regression and Experimental Design

Correlation is the gateway to more advanced statistical tools. Once you've established that a linear relationship exists, the natural next step is linear regression, which finds the best-fit line and lets you make predictions. And if you want to establish causation, you'll need to learn about experimental design — specifically, randomized controlled experiments. The table below compares what correlation can tell you versus what these more advanced methods add.

FeatureCorrelation (r)Linear RegressionControlled Experiment
What it tells youStrength and direction of a linear associationEquation of best-fit line; predicted values of yWhether changing x actually causes a change in y
Can predict?No — only measures associationYes — within the range of observed dataYes — with strong causal backing
Establishes causation?NoNo (uses observational data)Yes (with random assignment)
Type of studyObservationalObservational or experimentalExperimental

In future courses (like AP Statistics or college-level statistics), you'll explore regression analysis in depth. You'll also study how researchers design experiments with random assignment, control groups, and blinding to eliminate lurking variables and finally answer the question of cause and effect. For now, just remember: correlation is the essential first step — it identifies patterns worth investigating further.

Practice Problems

PROBLEM 1CONCEPTUAL
A researcher finds a correlation of r = −0.85 between the number of hours students spend on social media per day and their GPA. Describe the direction and strength of this correlation. Does this prove that social media use lowers GPA? Explain your reasoning.
PROBLEM 2BASIC CALCULATION
If the correlation between two variables is r = 0.60, calculate r² and explain what this value means in context.
PROBLEM 3INTERMEDIATE
A study finds that cities with more police officers also tend to have higher crime rates (r = 0.72). A politician concludes that hiring more police increases crime. Identify the flaw in this reasoning and propose at least one lurking variable that could explain the correlation.
PROBLEM 4APPLIED
A fitness app tracks data from 200 users and finds that daily step count and resting heart rate have a correlation of r = −0.45. The app advertises: 'Walking more will lower your heart rate!' Is this claim justified by the data? What kind of study would be needed to support it?
PROBLEM 5CRITICAL THINKING
Suppose you plot the data for a relationship and see the points form a perfect U-shape (parabola) with the lowest point in the middle. When you calculate r, you get approximately 0.05. Explain why r can be misleadingly low in this situation. What does this tell you about the limitations of the correlation coefficient?

Lesson Summary

The correlation coefficient (r) is a value between −1 and +1 that measures the strength and direction of a linear relationship between two quantitative variables. Values close to +1 indicate a strong positive association, values close to −1 indicate a strong negative association, and values near 0 suggest little or no linear pattern. The coefficient of determination (r²) tells you the percentage of variation in y that is explained by the linear relationship with x.

The most critical takeaway from this lesson is that correlation does not imply causation. A strong correlation can result from lurking variables, reverse causation, or pure coincidence. Only a randomized controlled experiment with proper controls can establish a cause-and-effect relationship. Always examine a scatterplot before interpreting r, since r only captures linear patterns and can miss strong non-linear relationships entirely.

Varsity Tutors • Math 2 • Interpreting Correlation