A dimensionless statistic r that measures linear association between two random variables by dividing their covariance by the product of their standard deviations; r takes values in the interval [-1,1].

Principle

Standardizing each variable removes units so that the expected product of standardized deviations quantifies linear dependence on a unitless normalized scale.

Demonstration

Two identical standardized signals have r = 1; for many pairs of independent zero-mean variables the expected r is zero, reflecting absence of linear association.

Misapplication

Interpreting a nonzero r as proof of causal influence, or using r to rule out dependence that is nonlinear or rank-based, are common misuse cases.

Consequence

Provides a simple, comparable measure of linear association used in diagnostics for linear models, screening of predictors, and as the basis for parametric tests when assumptions hold.

Reversal

Rank-based correlation measures operate on orderings rather than raw values and thus detect monotonic relationships that Pearson's r may miss while trading off sensitivity to linear scaling.

Boundary

Requires finite second moments and is undefined if either variable has zero variance; it is sensitive to outliers and does not capture nonlinear dependence structures.

Semantic Tension

Balances interpretability and comparability against blindness to many dependence forms; it is easy to report and compare but can understate association when relationships are nonlinear.

Synthesis

Pearson r is the normalized expected product of standardized deviations that quantifies the strength and sign of linear relation between two variables on a bounded scale.