covariance-and-correlation

Animated scatterplots show how covariance aggregates the signed products (X-EX)(Y-EY), and how correlation standardizes covariance by sigmaX*sigmaY to produce a unitless value in [-1,1]. The visualization cycles through negative correlation, positive correlation, and a zero-covariance-but-dependent (curved) case to emphasize that Cov=0 does not imply independence.

canvasclick to interact

t=0s

practical uses

01.Detecting linear relationships between features in data analysis and ML feature selection
02.Portfolio risk/hedging: understanding co-movement of asset returns (covariance matrix)
03.Checking model assumptions and diagnosing multicollinearity using correlation

technical notes

Generates deterministic pseudo-random samples. Uses correlated Gaussian construction Y=rho*U+sqrt(1-rho^2)*V and a dependent zero-covariance example Y=X^2-1. Covariance contributions are visualized via per-point brightness and small deviation-rectangle overlays. Responsive scaling via scale=min(w,h)/260 and grid snapping for a blocky aesthetic.

← game-theory-introduction linear-equations →