Covariance carries the combined units of two variables, while correlation is standardized. The standard deviations in a correlation denominator must be nonzero.
Where this helps
Describing how two quantities vary together.
Before you start
Try the examples to see what you’ll practice. Check the answer form and any stated conditions.
Suppose (X, Y) is equally likely to be (1, 2) or (3, 4).
E[X] = 2, E[Y] = 3, and E[XY] = (2 + 12)/2 = 7.
Cov(X, Y) = 7 − 2 × 3 = 1.
Use Cov(X, Y) = E[XY] − E[X]E[Y]. E means the probability-weighted average. Multiply paired values before averaging XY. The required expectations must be finite.
Multiply paired deviations
The paired observations are (1, 2), (3, 6), and (5, 4). Their means are x̄ = 3 and ȳ = 4.
Sum of deviation products = (−2)(−2) + (0)(2) + (2)(0) = 4.
Sample covariance = 4/(3 − 1) = 2.
Use sₓᵧ = Σ(xᵢ − x̄)(yᵢ − ȳ)/(n − 1), with n ≥ 2 pairs. Keep each x paired with its own y and divide by n − 1.
Scale covariance by both spreads
If Cov(X, Y) = 3, σₓ = 2, and σᵧ = 3, then:
ρₓᵧ = 3/(2 × 3) = 1/2.
Use ρₓᵧ = Cov(X, Y)/(σₓσᵧ). The symbols σₓ and σᵧ are the population standard deviations. Dividing by both removes the units. Both must be finite and strictly positive. If either is zero, this correlation is undefined.
Try a few examples
Real practice. Real progress.
Example 1 of 3
complete the population correlation formula from covariance
X and Y have finite positive standard deviations and a defined covariance.