What r value is considered a strong correlation?

This depends on your field. In physics and engineering, r > 0.99 might be expected. In social sciences, r > 0.50 is often considered strong. In psychology and education research, r > 0.30 is sometimes considered practically meaningful. A general rule of thumb: |r| 0.7 is strong. Always interpret r in the context of what is achievable and meaningful in your specific domain.

Does a high correlation mean one variable causes the other?

No. Correlation measures association, not causation. Three explanations always exist for a non-zero correlation: (1) X causes Y, (2) Y causes X, or (3) a third variable Z causes both X and Y (confounding). To establish causation, you need a randomised controlled experiment, a natural experiment, or careful causal analysis (e.g., instrumental variables, difference-in-differences). Correlation alone — no matter how high — cannot prove causation.

What is R² and how does it relate to r?

R² (the coefficient of determination) equals r². It represents the proportion of variance in one variable explained by the other. If r = 0.70, then R² = 0.49 — the two variables share 49% of their variance. R² is more intuitive for understanding practical effect size: r = 0.30 sounds small but only R² = 0.09 (9% of variance shared). Reporting R² alongside r gives readers a better sense of practical importance.

What is the difference between Pearson and Spearman correlation?

Pearson r measures linear association between two continuous variables and is sensitive to outliers. Spearman ρ measures monotonic association (a consistently increasing or decreasing relationship, not necessarily linear) and is based on ranks — making it robust to outliers and appropriate for ordinal data. Use Pearson when data is continuous and roughly normally distributed with no extreme outliers. Use Spearman when data is ordinal, non-normal, or contains outliers.

How do I test if a correlation is statistically significant?

Use a t-test: t = r × √(n−2) / √(1−r²), with df = n−2. If |t| exceeds the critical value for your alpha level (e.g., 1.96 for α = 0.05, two-tailed, with large n), the correlation is statistically significant — meaning it is unlikely to be due to chance if the null hypothesis (ρ = 0) is true. For small samples, you need a larger |r| to achieve significance: at n = 10, you need |r| > 0.632 for significance at α = 0.05.

Can I compare correlation coefficients from different studies?

With caution. Correlation coefficients are directly comparable only when studies use the same variables measured the same way in similar populations. Restriction of range, measurement error, and sample composition all affect the observed r. Fisher's z transformation (z = 0.5 × ln[(1+r)/(1−r)]) is used to test whether two correlations from different studies are significantly different from each other, accounting for the fact that sampling distributions of r are skewed near ±1.

Correlation Explained — What It Means and How to Interpret It

Name: Correlation Explained — Pearson r, Interpretation & Examples
Availability: OnlineOnly
Author: CalcMulti Editorial Team

By CalcMulti Editorial Team·Updated: February 2026·8 min read

Correlation measures the strength and direction of the linear relationship between two variables. The most commonly used measure is Pearson's correlation coefficient (r), which ranges from −1 to +1. A value of +1 means a perfect positive linear relationship; −1 means a perfect negative linear relationship; 0 means no linear relationship.

Understanding correlation is one of the most important and most misused skills in data analysis. This guide explains what correlation measures, how to calculate it, how to interpret r values, and — critically — why correlation does not imply causation.

Formula

r = Σ[(xᵢ − x̄)(yᵢ − ȳ)] / [(n−1) · sₓ · sᵧ]

What Correlation Actually Measures

Pearson r measures how well the relationship between two variables can be described by a straight line. When both variables tend to increase together, r is positive. When one tends to increase as the other decreases, r is negative. When there is no linear pattern, r is near 0.

Importantly, r measures only linear relationships. Two variables can have a strong curved (non-linear) relationship and still produce r ≈ 0. For example, the relationship between stress and performance forms an inverted-U curve (Yerkes-Dodson law) — at very low and very high stress, performance is poor; at moderate stress, performance peaks. Pearson r would be near 0 for this data even though the relationship is strong and predictable.

r is also dimensionless — it does not change if you rescale or shift the variables. The correlation between height in cm and weight in kg is the same as the correlation between height in inches and weight in pounds.

How to Interpret Correlation Coefficients

The sign of r indicates direction: positive r means variables increase together; negative r means one increases as the other decreases.

The magnitude of |r| indicates strength. Common rule of thumb (varies by field):

\|r\| Range	Strength	Real-World Example
0.00 – 0.19	Negligible / no relationship	Shoe size and intelligence
0.20 – 0.39	Weak	Hours of TV watched and GPA (r ≈ −0.25)
0.40 – 0.59	Moderate	Sleep duration and productivity (r ≈ 0.45)
0.60 – 0.79	Strong	Height and weight (r ≈ 0.70)
0.80 – 1.00	Very strong	Air temperature and ice cream sales (r ≈ 0.85)

Worked Example — Study Hours vs Exam Score

Five students — study hours (x): 2, 4, 6, 8, 10; exam scores (y): 55, 65, 70, 80, 90.

x̄ = 6, ȳ = 72. sₓ = 3.16, sᵧ = 13.5.

Deviations: (2−6)(55−72) = (−4)(−17) = 68 | (4−6)(65−72) = (−2)(−7) = 14 | (6−6)(70−72) = (0)(−2) = 0 | (8−6)(80−72) = (2)(8) = 16 | (10−6)(90−72) = (4)(18) = 72.

Sum of cross-products = 68 + 14 + 0 + 16 + 72 = 170.

r = 170 / [(5−1) × 3.16 × 13.5] = 170 / [4 × 42.66] = 170 / 170.64 ≈ 0.996.

Interpretation: r = 0.996 indicates a near-perfect positive linear relationship between study hours and exam score. As study hours increase by 1, exam score increases by approximately 4.3 points on average.

Types of Correlation Coefficients

Pearson r is the most common but is not always appropriate. Here are the main correlation measures and when to use each.

Correlation Type	Use When	Data Type	Key Property
Pearson r	Both variables are continuous and approximately linear	Continuous	Sensitive to outliers; assumes normality
Spearman's ρ (rho)	Ordinal data or non-linear but monotonic relationship	Ordinal or continuous	Based on ranks; robust to outliers
Kendall's τ (tau)	Small sample or many tied ranks	Ordinal	More conservative than Spearman; better for small n
Point-biserial	One binary variable, one continuous	Binary + continuous	Special case of Pearson r
Phi coefficient	Both variables are binary (0/1)	Binary	Special case of Pearson r for 2×2 tables

Common Mistakes in Correlation Analysis

1. Confusing correlation with causation. r = 0.85 between ice cream sales and drowning deaths is not because ice cream causes drowning — both are caused by hot weather (a confounding variable). Always consider alternative explanations before inferring causation from correlation.

2. Ignoring non-linear relationships. r = 0 does not mean no relationship — it means no linear relationship. Always plot your data (scatterplot) before reporting r. A curved or U-shaped pattern can have r ≈ 0.

3. Mistaking statistical significance for practical significance. With n = 1,000, a correlation of r = 0.10 is statistically significant (p < 0.001) but explains only 1% of the variance (R² = 0.01). Always report R² alongside r for practical context.

4. Correlation restricted range. If you only study students with high GPA, the correlation between study hours and GPA will be smaller (possibly near 0) than in the full student population. Restricting the range of a variable attenuates the observed correlation.

5. Assuming linearity without checking. Pearson r assumes the underlying relationship is linear. Use a scatterplot to confirm linearity before computing r.

Related Calculators

Correlation Calculator

Calculate Pearson r with significance test

Correlation vs Causation

Why correlation does not imply causation

Regression Analysis Explained

Using correlation to predict outcomes

Linear Regression Calculator

Calculate slope, intercept, and R²

Scatter Plot Guide

Statistics formulas reference

Statistics Hub

All statistics calculators & guides

← Back to Statistics Hub

Frequently Asked Questions

Educational use only. Content is based on publicly documented mathematical formulas and reviewed for accuracy by the CalcMulti Editorial Team. Last updated: February 2026.