Correlation Coefficient Calculator

The standard Pearson correlation coefficient (often denoted as r) evaluates how closely two continuous variables follow a linear path together. It is calculated by dividing the covariance of the two variables by the product of their standard deviations. For binary classifications or contingency tables, specialized variants like the Matthews correlation coefficient are used instead. The Matthews formula evaluates True Positives (TP), True Negatives (TN), False Positives (FP), and False Negatives (FN) using the equation: MCC = (TP × TN - FP × FN) / sqrt((TP + FP)(TP + FN)(TN + FP)(TN + FN)), yielding a balanced measure that performs reliably even with heavily skewed classes.

Imagine you are analyzing a medical diagnostic test for a rare condition. Out of your sample dataset, your classification model yields the following confusion matrix counts: True Positives (TP) = 50, True Negatives (TN) = 400, False Positives (FP) = 20, and False Negatives (FN) = 30. First, calculate the denominator product component: (50 + 20) × (50 + 30) × (400 + 20) × (400 + 30) = 70 × 80 × 420 × 430 = 1,010,880,000. Next, compute the numerator: (50 × 400) - (20 × 30) = 20,000 - 600 = 19,400. Finally, divide the numerator by the square root of the denominator: 19,400 / sqrt(1,010,880,000) = 19,400 / 31,794.34 ≈ 0.61. This strong positive score indicates your diagnostic model possesses a reliable predictive capability.

Always remember that correlation does not imply causation; two variables moving together may be driven by an unmeasured third confounding factor. Additionally, standard correlation coefficients only capture linear relationships, meaning complex quadratic or exponential trends might register a near-zero correlation despite being strongly related. Always visualize your data using scatter plots alongside numerical metrics to avoid misinterpretation.