Covariance

Also known as: Cov

Covariance $\mathrm{Cov}(X,Y)=E[XY]-\mu_X\mu_Y$ measures joint linear association. $\mathrm{Cov}(X,X)=\mathrm{Var}(X)$. Scaling: $\mathrm{Cov}(aX,bY)=ab\,\mathrm{Cov}(X,Y)$.

Independence $\Rightarrow$ $\mathrm{Cov}=0$; zero covariance (uncorrelated) does not imply independence. In simple OLS, $\hat\beta=\mathrm{Cov}(x,y)/\mathrm{Var}(x)$ (sample versions), and $\mathrm{Cov}(X,\varepsilon)=0$ is required for an unbiased slope.

← Back to CS Glossary