What does a positive correlation between two variables mean?
Choose an answer, then check it.
Data Science & AI Literacy · Foundations
Correlation
On this page 9 sections
In 30 seconds
correlation The tendency of two variables to change together in a predictable way, so that knowing one helps anticipate the other. Full entry → describes when two variables move together: as one changes, the other tends to change in a predictable way. positive correlation A pattern in which two variables tend to move in the same direction, both rising together or both falling together. Full entry → means they rise and fall together; negative means one rises while the other falls. Strength is how tightly the pattern holds, and a scatterplot A chart that shows the relationship between two variables by plotting one point per observation, positioned by its two values. Full entry → shows both at a glance. But correlation is not causation: ice cream sales and drowning rates both climb in summer because hot weather drives both. Data science models find correlations; deciding what they mean is human work.
Why this matters
Correlation is the most common relationship you will meet in data: claims that two things move together appear in news, marketing, health advice, and product recommendations. Understanding what correlation can and cannot tell you protects you from the most frequent error in reading data, mistaking a pattern for a cause. It matters academically because later topics, prediction, machine learning, and evaluating models, all build on the idea that patterns in data can be learned and used. It matters practically because AI systems constantly surface correlations, and the person who can ask what else could explain this pattern is the person who reads them correctly.
The college version
What correlation means
Correlation is about pairs. Whenever a dataset records two variables on the same cases, a student's height and weight, a city's temperature and its ice cream sales, the two variables can be studied together as a pair. Correlation describes the way such pairs move together: when one variable changes, the other tends to change in a predictable way. "Tends" matters, because real relationships are rarely perfect. Height and weight are correlated, yet two people of exactly the same height can weigh different amounts. Correlation is a description of a tendency in data, not a promise about any single case.
Positive and negative correlation
A positive correlation means the two variables move in the same direction: when one rises, the other tends to rise, and when one falls, the other tends to fall. Early in the growing season, the more weeks a gardener has watered a row of tomato plants, the taller the plants tend to be, so watering and plant height are positively correlated. A negative correlation A pattern in which one variable tends to rise while the other falls, so the two move in opposite directions. Full entry → means they move in opposite directions: when one rises, the other tends to fall. At a campus drink stand, hot drink sales climb as the outdoor temperature drops, so temperature and hot drink sales are negatively correlated. Both kinds are real patterns; they differ only in direction. And "tends" still applies: one slow afternoon can sell fewer hot drinks on a cold day without breaking the overall pattern.
Strong and weak
Strength is how closely the pattern holds. A strong correlation is a tight pattern: the points on the chart cluster close to a clear trend, so knowing one variable narrows the other down considerably. A weak correlation is a loose one: the points scatter widely around the trend, and knowing one variable tells you little about the other. Strength and direction are separate properties, and a strong negative correlation is every bit as real a pattern as a strong positive one. Correlation is often summarized on a scale from -1 to +1: the ends are perfect patterns, one in each direction, and zero means no straight-line pattern. The measure tracks straight-line patterns, so a strong curved pattern, such as a U shape, can look weak by this measure, which is one reason to always look at the chart rather than trust a number alone.
Seeing correlation with scatterplots
The standard way to see a correlation is a scatterplot: one point per observation, positioned by its two values, one variable on each axis. A rising cloud of points shows a positive correlation; a falling cloud shows a negative one; and how tightly the points cluster shows the strength. The scatterplot keeps the pairs together, which is exactly what correlation is about: it is the pairing, not the two variables separately, that reveals the relationship. How charts are designed and read belongs to the visualization topic; here it is enough to know that the scatterplot is the natural home of correlation.
Correlation is not causation
Here is the core lesson. Ice cream sales and drowning rates both rise in summer. The two variables are genuinely correlated. Does ice cream cause drowning? No. Hot weather is a third variable An unmeasured factor that drives both of the correlated variables, creating a pattern neither one causes; also called a confounding variable. Full entry → that drives both: warm days send people to the beach and to the ice cream stand, so both numbers climb together without one causing the other. This is the third-variable problem. It is so common that statisticians state the rule bluntly: correlation does not imply causation. The U.S. government's own statistical handbook puts it even more strongly: no statistical procedure, scatterplot included, proves cause and effect; such proof is beyond the realm of statistics. Correlation describes what happens together; it never says why.
Other explanations: reverse causation and coincidence
Any correlation between X and Y has at least four candidate explanations. One: X causes Y. Two: a third variable causes both. Three: reverse causation The situation in which the apparent direction of a relationship is backward, so the second variable actually influences the first. Full entry →, meaning the direction is backward and Y actually influences X. A hospital may find that patients who arrive by ambulance tend to have worse outcomes; the likely story is not that ambulances worsen patients, but that sicker patients are more likely to need an ambulance. Four: coincidence, a chance alignment of two unrelated numbers, such as a town's starling population and its county's streaming subscriptions drifting upward in the same years. Listing the candidates is the first step in checking any claim that two things move together.
Why this matters in data science and AI
Machine learning models are pattern finders: they learn correlations from data and use them to do useful work, and what models do with correlations is the prediction topic. But a model that finds a correlation does not know why the correlation exists. Whether the pattern reflects a cause, a shared third variable, reverse causation, or chance is a question for people: the analyst who inspects the data, the domain expert who understands the process, and the designer who decides what the model may be used for. Correlation literacy is therefore not a math skill but a thinking skill, and it sits at the center of data science.
Eli explains
The same idea, in plain words
Explain it like I’m 10
Correlation is when two variables move together: when one changes, the other tends to change in a predictable way. If they rise and fall together, that is a positive correlation; if one rises while the other falls, that is negative. A correlation can be strong, meaning the pattern holds tightly, or weak, meaning it barely holds at all. A scatterplot, one dot per observation, shows both direction and strength at a glance. The catch is that moving together is not the same as causing. Ice cream sales and drowning rates both climb in summer, not because ice cream causes drowning but because hot weather drives both. Data science models find these patterns; people decide what they mean.
Picture it like this
Picture two friends walking side by side. When one speeds up, the other usually speeds up; when one stops to look in a shop window, the other stops too. Watch them long enough and you could guess one from the other. That is correlation: the two move together, and the pattern is tight or loose depending on how in step they are. But watching them never tells you who sets the pace. One friend may be leading, or both may be hurrying toward the same bus stop.
Where the picture stops working
The friends can copy each other on purpose; data points cannot decide anything, so a correlation in data is only a pattern, never an intention. Two friends walking together are obviously connected, but two variables can move together for no reason at all, by coincidence, which the walking picture hides. And a perfect match in step tells you nothing about who is taller: correlation shows how two things move together, not how strongly one changes the other.
Worked example
A small-town café tracked two numbers each week for a year: cups of coffee sold and pages printed at the public library next door. Both climbed in exam weeks and dipped in quiet weeks, and the points on the scatterplot clustered along a rising trend. That is a positive correlation, and a fairly strong one. Does coffee cause people to print more pages? Almost certainly not. A third variable, the exam calendar, drives both: students study more, so they drink more coffee and print more notes. Could printing pages cause coffee drinking? Only if the library were the café's only customer, which it is not. The pattern is real; the story behind it still needs a human to figure out.
Key takeaway
Correlation describes how two variables move together, the direction and the strength of the pattern, but never why. A correlation may reflect a cause, a shared third variable, reverse causation, or coincidence; telling these apart is human judgment, not a statistic.
Quick check
3 questions here, of 5 in this lesson’s practice set. Answers stay hidden until you check.
Ice cream sales and drowning rates both rise during summer. Why is it wrong to conclude that ice cream causes drownings?
Choose an answer, then check it.
A food truck records its sales for a year and finds that as the outside temperature drops, sales of hot soup rise. What kind of correlation is this?
Choose an answer, then check it.
Study tools & related lessonsYou’ll learn to · Common mistakes · Easily confused · Key vocabulary · Related
You’ll learn to
- Define correlation as the tendency of two variables to change together in a predictable way.
- Distinguish positive from negative correlation and give an original example of each.
- Explain how strong and weak correlations differ in how closely the pattern holds.
- Describe how a scatterplot reveals the direction and strength of a relationship between two variables.
- Explain why correlation does not imply causation, naming the third-variable, reverse-causation, and coincidence alternatives.
- Apply the correlation-is-not-causation check to a claim made about data.
Common mistakes
Concluding that if two variables move together, one must cause the other.
A shared pattern can come from a third variable, reverse causation, or coincidence. Correlation describes what happens together, not why.
Treating a negative correlation as if it meant the variables have no relationship.
Negative correlation is a real relationship: the variables move in opposite directions, like temperature and hot drink sales.
Reading a strong correlation as a strong effect.
Strength describes how tightly the pattern holds. It does not say how much one variable changes when the other does, and it does not say which variable, if either, is doing the causing.
Assuming a weak correlation means nothing is going on.
The standard measure captures straight-line patterns, so a strong curved relationship can show up as a weak correlation. Check the scatterplot, not just the number.
Easily confused
Positive correlation vs. Negative correlation
Both describe a consistent pattern, but positive means the variables move in the same direction while negative means they move in opposite directions.
Correlation vs. Causation
Correlation is an observed pattern of two variables moving together; causation is a claim that one brings about the other, a claim no correlation alone can support.
Strong correlation vs. Weak correlation
Both have a clear direction; strong means the pattern holds tightly with points clustered together, weak means it barely holds with points scattered widely.
Key vocabulary
- correlation
- The tendency of two variables to change together in a predictable way, so that knowing one helps anticipate the other.
- positive correlation
- A pattern in which two variables tend to move in the same direction, both rising together or both falling together.
- negative correlation
- A pattern in which one variable tends to rise while the other falls, so the two move in opposite directions.
- scatterplot
- A chart that shows the relationship between two variables by plotting one point per observation, positioned by its two values.
- linear relationship
- A relationship whose points cluster along a straight line; the kind of pattern correlation measures best.
- third variable
- An unmeasured factor that drives both of the correlated variables, creating a pattern neither one causes; also called a confounding variable.
- reverse causation
- The situation in which the apparent direction of a relationship is backward, so the second variable actually influences the first.
- bivariate data
- Observations that record two variables on each case, such as a person's height and weight, kept together so the pairing is not lost.
Sources & references
- NIST/SEMATECH e-Handbook of Statistical Methods — 1.3.3.26. Scatter Plot — NIST / SEMATECH (U.S. Department of Commerce)
- OpenIntro — Introduction to Modern Statistics (2e), Section 7.1: Fitting a line, residuals, and correlation — OpenIntro (openintro.org)
- Online Statistics Education: An Interactive Multimedia Course of Study — Describing Bivariate Data: Introduction — Rice University (Rudy Guerra and David M. Lane)
- Introductory Statistics, Section 12.2: Scatter Plots — OpenStax (Rice University)
- Correlation does not imply causation — Wikipedia
EliExplains lessons are original prose written from the open, credible references above. See Copyright & Licensing.
Researched 2026-08-21
Educational content only. It is not medical, legal or professional advice. Found an error? Tell us.