Correlation Coefficient Calculator

Pearson's correlation coefficient (r) measures the strength and direction of a LINEAR relationship between two variables. It's the most fundamental statistic for understanding bivariate relationships. **The Formula**: r = (nΣxy - ΣxΣy) / √[(nΣx² - (Σx)²)(nΣy² - (Σy)²)] Or equivalently: r = Σ[(x - x̄)(y - ȳ)] / √[Σ(x - x̄)² × Σ(y - ȳ)²] **Range and Interpretation**: | r | Strength | |---|----------| | +1.0 | Perfect positive | | +0.8 to +1.0 | Strong positive | | +0.5 to +0.8 | Moderate positive | | +0.3 to +0.5 | Weak positive | | 0 to +0.3 | Very weak/no relationship | | 0 | No linear relationship | | -0.3 to 0 | Very weak negative | | -0.5 to -0.3 | Weak negative | | -0.8 to -0.5 | Moderate negative | | -1.0 to -0.8 | Strong negative | | -1.0 | Perfect negative | **Examples of Correlations**: - **Height vs weight**: r ≈ 0.7 (strong positive) - **Education vs income**: r ≈ 0.5 (moderate positive) - **SAT score vs GPA**: r ≈ 0.4 (moderate positive) - **Temperature vs ice cream sales**: r ≈ 0.8 (strong positive) - **Cigarette smoking vs lung cancer**: r ≈ 0.7 (strong positive) - **Hours of TV vs grades (children)**: r ≈ -0.3 (weak negative) **Coefficient of Determination (r²)**: The square of the correlation coefficient (r²) represents the proportion of variance in y explained by x. - r = 0.5 → r² = 0.25 (25% of variance explained) - r = 0.7 → r² = 0.49 (49% of variance explained) - r = 0.9 → r² = 0.81 (81% of variance explained) This is why correlations of 0.3 are 'weak' — they only explain 9% of variance. **Critical Insight: Correlation ≠ Causation** A strong correlation does NOT prove causation. There are several possibilities: 1. **X causes Y**: Smoking causes lung cancer 2. **Y causes X**: Reverse causation 3. **Z causes both**: Confounding variable (lurking variable) 4. **Coincidence**: Random correlation Classic example: Ice cream sales and drowning deaths are highly correlated. But ice cream doesn't cause drowning — both increase in summer (the lurking variable). **Famous Spurious Correlations**: - US spending on science vs suicides by hanging (r ≈ 0.99) - Per capita cheese consumption vs civil engineering doctorates (r ≈ 0.96) - Number of pirates vs global warming (inverse correlation) These are coincidences, not causal relationships. **Assumptions of Pearson's r**: 1. **Linear relationship**: Doesn't capture curves 2. **Both variables continuous** and approximately normal 3. **No extreme outliers** 4. **Homoscedasticity**: Equal variance across x values Violations require alternative methods: - **Spearman's rho**: For ordinal data or non-linear monotonic - **Kendall's tau**: Similar to Spearman, less sensitive to ties - **Mutual information**: For non-linear, non-monotonic **Sample Size Effects**: With small samples, even strong correlations may not be statistically significant. Required sample sizes for significance: | r | n needed for p < 0.05 | |---|----------------------| | 0.3 | ~44 | | 0.5 | ~16 | | 0.7 | ~9 | | 0.9 | ~5 | **Hypothesis Testing**: To test if r is significantly different from 0: t = r × √(n-2) / √(1-r²) Compare to t-distribution with n-2 degrees of freedom. **Computing by Hand** (Example): Given 5 data points: (1,2), (2,4), (3,5), (4,7), (5,8) n = 5 Σx = 15, Σy = 26 Σxy = 1×2 + 2×4 + 3×5 + 4×7 + 5×8 = 2 + 8 + 15 + 28 + 40 = 93 Σx² = 1 + 4 + 9 + 16 + 25 = 55 Σy² = 4 + 16 + 25 + 49 + 64 = 158 r = (5×93 - 15×26) / √[(5×55 - 15²)(5×158 - 26²)] r = (465 - 390) / √[(275 - 225)(790 - 676)] r = 75 / √(50 × 114) r = 75 / √5700 r = 75 / 75.50 r = 0.993 **Practical Tips**: 1. **Always plot the data first** — see if relationship looks linear 2. **Check for outliers** that may distort r 3. **Use larger samples** for stable estimates 4. **Don't overinterpret** small correlations 5. **Report r² along with r** for context 6. **Test for significance** before drawing conclusions