You've seen it on a test. " And it is. Day to day, you've probably stared at a list of numbers and thought, "One of these has to be wrong. Even so, strict ones. Because the correlation coefficient r doesn't just take any value it feels like. On the flip side, you've seen it in a textbook. It has rules. And breaking them is a fast way to lose points, lose credibility, or lose your mind during an exam And it works..
Some disagree here. Fair enough.
What Is a Correlation Coefficient, Really
Let's get this straight. The correlation coefficient, usually called r, measures the strength and direction of a linear relationship between two variables. Now, that's the textbook line. But here's what it actually means in practice: it tells you how well a straight line fits the data when you plot the two variables against each other That's the whole idea..
r can be positive or negative. Positive means as one variable goes up, the other tends to go up too. Negative means as one goes up, the other tends to go down. And the closer r is to 1 (or -1), the tighter that relationship is. A value near zero means there's basically no linear relationship at all The details matter here..
Here's what most people miss. r is not about causation. It never was. Day to day, it's about association. You can have two variables that move together perfectly and still not have one causing the other. But that's a different conversation. Right now, we're focused on something more basic. What values can r actually take?
Why This Matters More Than You Think
I know, it sounds like a tiny detail. On top of that, a range. So who cares? But here's the thing — understanding the possible values of r is the foundation for everything else you'll do with correlation. Even so, if you don't nail this, you'll misinterpret scatterplots. Now, you'll pick the wrong answer on a test. You'll write a report that makes a statistician cringe Not complicated — just consistent..
And in real talk, this shows up everywhere. Which means in sports analytics. Consider this: in psychology research. In epidemiology. In marketing data. So anywhere two things are measured together, someone's calculating r. If you can't spot an impossible value, you can't spot bad data. And bad data leads to bad decisions Not complicated — just consistent..
This is where a lot of people lose the thread.
The short version is this: r is bounded. That said, nothing outside that range is valid. It lives between -1 and 1. Nothing.
How It Works: The Possible Values of r
r is calculated using a formula that looks intimidating but really just standardizes the relationship between two variables. The formula is:
r = Σ[(x - x̄)(y - ȳ)] / √[Σ(x - x̄)² * Σ(y - ȳ)²]
Don't panic. Still, you don't need to memorize it. What you need to understand is what it produces.
The range is always -1 to 1
r can be exactly -1. That means a perfect negative linear relationship. Every point falls on a straight line that slopes downward. Similarly, r can be exactly 1, which means a perfect positive linear relationship. Every point falls on a line that slopes upward.
Anything between -1 and 1 is possible. Because of that, -0. 87. Think about it: 0. 34. -0.12. That's why 0. Because of that, 99. All valid.
And here's a key point: r can be zero. Even so, the points are scattered with no apparent trend. That means no linear relationship. This is completely legitimate.
Why can't r be outside -1 to 1?
Because the formula is built on the Cauchy-Schwarz inequality. In plain English, that means the numerator (the covariance part) can never be larger in absolute value than the denominator (the product of the standard deviations). The math guarantees the ratio stays within bounds Small thing, real impact..
You can think of it like this. Here's the thing — r is a ratio. And ratios that come from this specific calculation are capped. There's no way to squeeze a value of 1.Also, 5 or -2 out of it if the data is real. Still, if you ever calculate an r outside that range, something went wrong. Your data, your formula, your calculator — something And that's really what it comes down to. Nothing fancy..
What does a possible r value look like in context
Say you're looking at height and weight in a group of adults. Because of that, you'd expect a positive correlation. Maybe r = 0.73. That's reasonable. Strong, but not perfect. Now say someone claims the correlation between study hours and exam score is 1.Even so, 2. Practically speaking, that's not possible. Practically speaking, period. Or if someone says the correlation between age and marathon time is -1.5. Also impossible Worth keeping that in mind..
This is where the question "which of the following is not a possible r value" comes from. It's testing whether you know the boundaries.
Common Mistakes People Make
Honestly, this is the part most guides get wrong. They'll list the range and move on. But the mistakes are subtler than that Small thing, real impact..
Confusing r with r-squared
r is the correlation coefficient. r² is the coefficient of determination. r² is always positive, even if r is negative, because you're squaring it. So r² ranges from 0 to 1. Some people see a value like 0.81 and think, "That's a possible r." And sure, r could be 0.9, since 0.9² = 0.81. But r² itself is not r. Don't mix them up on a test Easy to understand, harder to ignore..
Thinking r can be greater than 1 if the relationship is "really strong"
No. That said, " The math doesn't allow it. 1, check your inputs. If your data seems to produce an r of 1.There's no such thing as "more than perfect.97 is already extremely strong. Now, a correlation of 0. Worth adding: strength doesn't push r past 1. You probably swapped a sign, or duplicated a data point, or used the wrong column Simple, but easy to overlook..
Assuming all values between -1 and 1 are equally likely
They're not. Now, in real datasets, you'll see a lot of values clustered toward zero. Strong correlations are less common than weak ones. Think about it: that's just how data tends to behave. But that doesn't change what's possible. Now, -0. Day to day, 99 is possible. So is 0.Which means 01. Both are valid.
Forgetting that r only measures linear relationships
This one trips people up in interpretation, not in the "possible value" question. But it's worth saying. r only captures linear association. You can have a perfect curved relationship and still get r close to zero. That doesn't mean there's no relationship. It means there's no linear one. The question of possible values is about the number itself, but context matters for what that number means.
Practical Tips: How to Spot an Impossible r Value Fast
Here's what I'd actually do if I were sitting in an exam and saw a list of options.
First, look for anything greater than 1 or less than -1. Done. In real terms, that's your answer. No calculation needed.
If all options are within the range, look for trick answers. Is one of them r² instead of r? Is one of them a p-value? Practically speaking, is one of them a slope from a regression? These get mixed in to test whether you're paying attention.
Not obvious, but once you see it — you'll see it everywhere.
Second, if you have time, do a quick mental check. That said, does the direction make sense with the variables? So if someone says the correlation between hours of sleep and feeling tired is positive, that's probably wrong. So naturally, more sleep should correlate with less tiredness. But that's about logic, not the range. Still, it's a good habit.
Third, remember that *
