Variance of Linear Transformations
You'll learn how multiplying and adding constants to a random variable affects its spread — a key tool for simplifying probability calculations.
The Core Idea
When you transform a random variable X by scaling it (multiplying by a constant a) or shifting it (adding a constant b), the variance changes in a specific, predictable way:
Var(aX + b) = a²Var(X)
Notice two critical facts:
- Scaling squares the effect: Multiplying
Xbyamultiplies the variance bya², not justa - Shifting does nothing: Adding a constant
bhas zero effect on variance
Why This Makes Sense
Think of variance as measuring how "spread out" your data is from the mean.
-
Shifting (
+b): If you add 10 to every value in a dataset, the values all move together. The distance between values stays the same, so the spread (variance) doesn't change. -
Scaling (
×a): If you double every value, the distances between points also double. But variance measures squared distances from the mean, so doubling creates four times the variance (2² = 4).
Quick Example
If Var(X) = 9:
Var(X + 5) = 9(shift by 5 → no change)Var(3X) = 3² × 9 = 9 × 9 = 81(triple the values → 9× the variance)Var(3X - 7) = 9 × 9 = 81(the -7 shift still has no effect)
This formula lets you quickly compute variance for transformed variables without recalculating from scratch — extremely useful when analyzing algorithm performance with scaled inputs.
Key Takeaway: Variance responds to scaling quadratically (a²) but ignores constant shifts entirely: Var(aX + b) = a²Var(X).