Pearson Correlation Coefficient Calculator
Enter two paired data sets and instantly get Pearson's correlation coefficient r, the coefficient of determination r², covariance and a plain-English interpretation of the relationship's strength and direction.
Calculate Pearson Correlation
Correlation Coefficient Formula
Where:
- \(r\) — Pearson correlation coefficient, always between \(-1\) and \(+1\)
- \(x_i, y_i\) — the paired data values
- \(\bar{x}, \bar{y}\) — the means of the two data sets
- \(n\) — the number of pairs
The numerator is the sum of cross-products of deviations (the covariance numerator); the denominator rescales it by the product of the spreads of X and Y, so the result becomes a dimensionless number. This is why r is unaffected by the units or scale of your data.
The calculator also reports the coefficient of determination \( r^2 \) and the covariance:
\[ r^2 = r \times r, \qquad \operatorname{cov}(X,Y) = \frac{\sum_{i=1}^{n} (x_i - \bar{x})(y_i - \bar{y})}{n - 1 \ \text{(sample)} \ \text{or} \ n \ \text{(population)}} \]
Correlation Coefficient – Calculation Example
Example: a researcher measures study hours (X) and test scores (Y) for eight students: X = 1, 2, 3, 4, 5, 6, 7, 8 and Y = 53, 61, 68, 74, 80, 85, 90, 95.
\[ \bar{x} = \frac{1+2+\cdots+8}{8} = 4.5, \qquad \bar{y} = \frac{53+61+\cdots+95}{8} = 75.75 \]
\[ \sum (x_i - \bar{x})(y_i - \bar{y}) = 248, \qquad \sum (x_i - \bar{x})^2 = 42, \qquad \sum (y_i - \bar{y})^2 = 1475.5 \]
\[ r = \frac{248}{\sqrt{42 \times 1475.5}} = \frac{248}{248.94} \approx 0.9962 \]
Result: r ≈ 0.9962, a very strong positive linear relationship. The coefficient of determination is r² ≈ 0.9925, meaning about 99.25% of the variation in test scores is explained by study hours. With the "Sample" covariance type, cov(X, Y) = 248 / 7 ≈ 35.43; with "Population", cov(X, Y) = 248 / 8 = 31. There are n = 8 pairs.
How to Use the Calculator
- Enter your first variable's values in the Data set X field. Separate numbers with commas, spaces, semicolons or new lines — whichever you prefer.
- Enter the matching values in Data set Y, in the same order as X. Each X value must be paired with exactly one Y value, and both sets must contain the same number of values.
- Choose the Covariance type: "Sample (divide by n − 1)" if your pairs are a sample of a larger population, or "Population (divide by n)" if you have data for the entire population. This choice affects only the reported covariance — the correlation coefficient r is identical either way.
- Read the results: Pearson's r, the coefficient of determination r², the covariance, the number of pairs n, and a plain-language interpretation of the relationship.
What Is the Pearson Correlation Coefficient?
The Pearson correlation coefficient (usually written as r) measures the strength and direction of a linear relationship between two numeric variables. It always lies between \(-1\) and \(+1\):
- +1 — a perfect positive linear relationship: every increase in X comes with a proportional increase in Y.
- 0 — no linear association between the variables.
- −1 — a perfect negative linear relationship: as X increases, Y decreases proportionally.
The sign tells you the direction of the relationship, while the absolute value tells you how tightly the data points cluster around a straight line. Because r is standardized, it does not depend on the units of measurement: the correlation between height and weight is the same whether you measure in kilograms and centimeters or pounds and inches.
The coefficient of determination r² answers a related question: how much of the variation in Y is explained by its linear relationship with X. For example, r = 0.7 gives r² = 0.49, so roughly 49% of the variation in Y is accounted for by the linear trend. This is often a more intuitive and honest way to report the practical size of a relationship.
The covariance reported alongside r is the unscaled version of the same idea. Covariance describes the direction of the joint variation but its magnitude depends on the units, which is exactly why the standardized r is usually the more useful statistic.
Interpreting the Result: A Rough Guide
The interpretation scale used by this calculator is a common descriptive convention. Treat the boundaries as approximate, not strict rules — context matters:
| |r| | Descriptive strength |
|---|---|
| 0.90 – 1.00 | Very strong |
| 0.70 – 0.89 | Strong |
| 0.50 – 0.69 | Moderate |
| 0.30 – 0.49 | Weak |
| 0.00 – 0.29 | Very weak to none |
Assumptions, Limitations and Practical Tips
- Linearity: Pearson's r only detects straight-line patterns. A relationship can be very strong but curved (for example, Y = X²) and still yield r close to 0. Always plot your data before drawing conclusions from r alone.
- Correlation is not causation: a high r shows that two variables move together, not that one causes the other. A hidden third variable may drive both.
- Outliers: r is sensitive to extreme values. A single unusual point can inflate or deflate the coefficient dramatically, especially with small samples.
- Paired data: the calculation requires that each X value is paired with one Y value (same source or unit). Two unrelated lists of numbers produce a meaningless r.
- Sample size: with very few pairs, even a large r can arise by chance. With n = 2 the coefficient is always exactly +1 or −1, which is why at least 2 pairs are required and more are strongly recommended.
- Continuous data: Pearson's r is designed for continuous, roughly numeric data. For ordinal (ranked) data, Spearman's rank correlation is usually the more appropriate choice.
Finally, remember the difference between the covariance options: the divisor (n − 1 versus n) changes the covariance value but never changes r or r², since the same factor appears in the numerator and denominator of the correlation formula.
Frequently Asked Questions
How do I calculate Pearson's correlation coefficient by hand, step by step?
First compute the means \( \bar{x} \) and \( \bar{y} \). Then, for each pair, find the deviations \( (x_i - \bar{x}) \) and \( (y_i - \bar{y}) \), and build three sums: the cross-products \( \sum (x_i - \bar{x})(y_i - \bar{y}) \), and the squared deviations \( \sum (x_i - \bar{x})^2 \) and \( \sum (y_i - \bar{y})^2 \). Finally, divide the cross-product sum by the square root of the product of the two squared-deviation sums. A spreadsheet table with one column per step keeps the arithmetic manageable.
Is my correlation coefficient statistically significant?
This calculator reports the size of r but not a p-value. As a rough check, you can compute \( t = r \sqrt{\frac{n-2}{1-r^2}} \) and compare it to a t-distribution with \( n - 2 \) degrees of freedom. With n = 30 pairs, |r| above roughly 0.36 is significant at the 0.05 level; with n = 100, |r| above about 0.20 suffices. Small samples need much larger correlations to be meaningful, so treat borderline results with caution.
Can Pearson's r ever be greater than 1 or less than −1?
No. Mathematically, r is always between −1 and +1, because the denominator \( \sqrt{\sum(x_i-\bar{x})^2 \cdot \sum(y_i-\bar{y})^2} \) can never be smaller than the numerator (this is the Cauchy–Schwarz inequality). If your own calculation produces a value outside this range, it signals a computational or data-entry error — most often mismatched pairs or a misplaced minus sign. Values like exactly +1 or −1 with only two pairs are mathematically forced, not evidence of a perfect relationship.
What's the difference between Pearson and Spearman correlation?
Pearson's r measures how well the data follow a straight line and assumes roughly continuous, interval-scaled variables. Spearman's rank correlation first converts values to ranks and then correlates those ranks, so it captures any consistently monotonic relationship — straight or curved — and is robust to outliers. Use Spearman for ordinal data (ratings, rankings) or when a scatterplot shows a curved but always-increasing or always-decreasing pattern. Use Pearson when linearity and roughly normal-ish data are reasonable.
Does converting units, like centimeters to inches, change the correlation?
No — Pearson's r is dimensionless and invariant to linear changes of scale. Multiplying or adding constants to either variable (converting currencies, shifting a baseline, rescaling scores) leaves r exactly the same, because the rescaling factor appears in both the numerator and the denominator and cancels out. The sign flips only if you multiply one variable by a negative number. The covariance, however, does change with units, which is one reason r is usually the preferred statistic.
Why does one outlier have such a big effect on my result?
Because r is built from sums of deviations, a single extreme point contributes disproportionately large cross-product terms. With small samples, one outlier can drag r from 0.8 down to near 0, or push a weak relationship to look strong. Check your data for typos and measurement errors first, then plot a scatterplot to see whether any point dominates. If the outlier is legitimate but distorts the picture, consider reporting Spearman's rank correlation or a robust alternative alongside Pearson's r.
What are the most common mistakes when computing a correlation?
The biggest one is unpaired or misordered data — the i-th X value must correspond to the i-th Y value, and both lists must be the same length. Other frequent errors include using r for curved relationships (where it understates the association), confusing r with r² when reporting "percent of variation explained," and treating a large r as proof of causation. Also remember that a constant data set makes r undefined, since its spread is zero.