Expected Value of Functions
What you'll learn: How to compute the expected value of a transformed random variable—like squaring it or plugging it into a formula—without listing every outcome manually.
The Core Idea
You already know how to compute the expected value of a random variable X. But what if you need the expected value of some function of X, like X² or 2X + 3?
The Law of the Unconscious Statistician (LOTUS) gives you a shortcut: you don't need to first figure out the probability distribution of the new random variable. Instead, you can compute E[g(X)] directly using the original distribution of X.
For Discrete Random Variables
If X is discrete with PMF P(X = x), then:
E[g(X)] = Σ g(x) · P(X = x)
Sum over all possible values of X, but apply your function g to each value before weighting by its probability.
Example
Suppose X represents the outcome of rolling a fair six-sided die. What's the expected value of X²?
- Instead of creating a whole new random variable "X-squared," just compute:
- E[X²] = (1² × 1/6) + (2² × 1/6) + (3² × 1/6) + (4² × 1/6) + (5² × 1/6) + (6² × 1/6)
- E[X²] = (1 + 4 + 9 + 16 + 25 + 36) / 6 = 91/6 ≈ 15.17
You applied the function (squaring) to each outcome, then weighted by probabilities.
Why It Matters in CS
This technique appears everywhere: analyzing randomized algorithms (expected runtime of T²), hashing collisions, and variance calculations all rely on computing expectations of functions.
Key Takeaway: Use the Law of the Unconscious Statistician to compute E[g(X)] directly by applying g to each value of X in the original probability distribution.