Pairwise vs Mutual Independence
What you'll learn: You'll discover through concrete examples that even when every pair of random variables is independent, all three together might not be mutually independent.
The Core Distinction
When working with three or more random variables, there's a subtle but crucial difference:
- Pairwise independence: Every pair of variables is independent (X and Y, X and Z, Y and Z)
- Mutual independence: The entire collection is independent together
Surprisingly, pairwise independence is weaker than mutual independence. Variables can be pairwise independent without being mutually independent.
A Classic Counterexample
Consider flipping two fair coins. Define three random variables:
- X = 1 if first coin is heads, 0 otherwise
- Y = 1 if second coin is heads, 0 otherwise
- Z = 1 if both coins match (both heads or both tails), 0 otherwise
Check pairwise independence:
- X and Y are independent (coin flips are independent) ✓
- X and Z: P(X=1, Z=1) = 1/4 = P(X=1)·P(Z=1) = (1/2)(1/2) ✓
- Y and Z: Similarly independent ✓
Check mutual independence: For mutual independence, we need P(X=x, Y=y, Z=z) = P(X=x)·P(Y=y)·P(Z=z) for all combinations.
But observe: P(X=1, Y=0, Z=1) = 0 (if coins differ, they can't match!)
However: P(X=1)·P(Y=0)·P(Z=1) = (1/2)(1/2)(1/2) = 1/8 ≠ 0
The joint distribution doesn't factor completely, so they're not mutually independent.
The Insight
Z depends on the combination of X and Y together, even though knowing just X (or just Y) alone tells you nothing about Z. Pairwise independence misses these higher-order relationships.
Key Takeaway: Pairwise independence of all pairs is necessary but not sufficient for mutual independence of three or more random variables—you must verify the joint distribution factors completely.