Correlation Coefficient — Pearson's r & Interpretation

The correlation coefficient usually means Pearson's correlation coefficient, written $r$. It measures the direction and strength of a linear relationship between two numerical variables.

If $r$ is positive, the variables tend to increase together. If $r$ is negative, one tends to decrease as the other increases. If $r$ is near $0$, Pearson's $r$ is saying there is little linear pattern, not necessarily no relationship at all.

Pearson's $r$ is most useful when the data come in pairs, both variables are numerical, and a straight-line trend is the pattern you want to summarize.

What The Correlation Coefficient Tells You

Pearson's $r$ is a standardized measure of how two variables vary together. For a sample of paired data, the formula is

$$r = \frac{\sum_{i=1}^n (x_i-\bar{x})(y_i-\bar{y})}{\sqrt{\sum_{i=1}^n (x_i-\bar{x})^2 \sum_{i=1}^n (y_i-\bar{y})^2}}$$

The numerator is positive when the variables tend to move in the same direction and negative when they tend to move in opposite directions. The denominator rescales that joint movement using the spread of each variable.

When Pearson's $r$ is defined, it must satisfy

$$-1 \le r \le 1$$

If one variable has no variation at all, the denominator becomes $0$, so Pearson's $r$ is undefined.

How To Interpret Positive, Negative, And Near-Zero Values

Start with the sign:

• $r > 0$: positive linear association
• $r < 0$: negative linear association
• $r = 0$: no linear association

Then look at the magnitude $|r|$. Values closer to $1$ mean the points stay closer to a straight-line pattern. Values closer to $0$ mean the linear pattern is weaker.

Be careful with labels like "weak," "moderate," or "strong." Those cutoffs depend on context. In one field, $r = 0.3$ may matter. In another, it may be too small to support a decision.

The safest habit is to read $r$ alongside a scatter plot. The number is a summary of the pattern you see; it should not replace the picture.

Worked Example: Calculating $r = 0.9$

Suppose the paired data are

$$(1,2),\ (2,3),\ (3,5),\ (4,4),\ (5,6)$$

First compute the means:

$$\bar{x} = \frac{1+2+3+4+5}{5} = 3$$ $$\bar{y} = \frac{2+3+5+4+6}{5} = 4$$

Now list the deviations from the means:

• For $x$: $-2, -1, 0, 1, 2$
• For $y$: $-2, -1, 1, 0, 2$

Multiply the paired deviations and add:

$$(-2)(-2) + (-1)(-1) + (0)(1) + (1)(0) + (2)(2) = 4 + 1 + 0 + 0 + 4 = 9$$

Now compute the two sums of squares:

$$\sum (x_i-\bar{x})^2 = 4+1+0+1+4 = 10$$ $$\sum (y_i-\bar{y})^2 = 4+1+1+0+4 = 10$$

So

$$r = \frac{9}{\sqrt{10 \cdot 10}} = \frac{9}{10} = 0.9$$

This tells you there is a strong positive linear association in this sample. As $x$ increases, $y$ usually increases too, and the points would sit fairly close to an upward-sloping line.

Common Mistakes When Interpreting Correlation

Treating Correlation As Causation

A high correlation does not prove that one variable causes the other. A third factor may influence both, or the relationship may be coincidental in the observed data.

Forgetting That Pearson's $r$ Is Linear

Pearson's $r$ only measures linear association well. A curved relationship can produce a small correlation even when the variables are clearly related.

Ignoring Outliers

One unusual point can change $r$ a lot. If the scatter plot has an outlier, the correlation may tell a misleading story about the overall pattern.

Using Pearson's $r$ When The Setup Does Not Fit

Pearson's $r$ is designed for paired numerical data and linear association. If one variable is categorical, or if the pattern is clearly curved, this coefficient may not answer the question you actually care about.

Overreading A Near-Zero Value

A value near $0$ means "little linear association," not "no relationship of any kind."

When Pearson's Correlation Coefficient Is Used

Pearson's $r$ is commonly used in statistics, science, economics, social research, and machine learning as a quick summary of paired numerical data. It is most useful when you want to know whether a straight-line pattern is present before moving to a model such as linear regression.

In practice, a scatter plot should come first. The coefficient is a summary, not a replacement for looking at the data.

Try A Similar Problem

Take a small data set you already understand, plot the points, and estimate whether the trend looks positive, negative, or unclear before calculating $r$. That quick comparison is one of the fastest ways to build intuition for what the correlation coefficient is actually saying.

If you want to go one step further, explore the same data with a simple linear regression line. That makes it easier to see how correlation and prediction are related, but not identical.