A topic in the Open Knowledge Graph — a free, open map of 15,290 topics and the order to learn them in.
Linear transformations are the workhorses of probability. If Y = aX + b, then E[Y] = aE[X] + b and Var(Y) = a²Var(X)—expectation is linear while variance scales quadratically and is unaffected by shifts. For sums of random variables, E[X + Y] = E[X] + E[Y] always holds, but Var(X + Y) = Var(X) + Var(Y) + 2Cov(X, Y), so independence simplifies the variance formula. These properties underpin standardization (converting any distribution to mean 0 and variance 1), the construction of confidence intervals, and the derivation of sampling distributions used throughout statistical inference.
Derive the rules algebraically from the definition of expectation, then verify them numerically with a simple example: roll a die, let X be the result, compute E[3X + 2] and Var(3X + 2) both by formula and by enumerating all outcomes.
Students frequently forget the squared coefficient in Var(aX) = a²Var(X) and write aVar(X) instead. Another common error is assuming Var(X + Y) = Var(X) + Var(Y) without checking independence—the covariance term is only zero when X and Y are uncorrelated.
From your study of expected value and variance, you know that E[X] measures the center of a distribution and Var(X) measures its spread. Linear transformations — operations of the form Y = aX + b — are the most common manipulations of random variables, and understanding how they affect the mean and variance is essential for everything from standardization to sampling distributions to confidence intervals.
The rules are clean but asymmetric. For the mean: E[aX + b] = aE[X] + b — expectation is perfectly linear. Scaling X by a scales the mean by a; shifting by b shifts the mean by b. This follows directly from the linearity of summation (or integration). For the variance: Var(aX + b) = a²Var(X) — the shift b disappears entirely, and the scale factor enters squared. The shift vanishes because adding a constant moves every value by the same amount, leaving the spread unchanged. The squaring arises because variance is defined as E[(X − μ)²], and scaling X by a scales each deviation by a, which after squaring gives a². The standard deviation, being the square root of variance, scales linearly: SD(aX + b) = |a| · SD(X).
For sums of random variables, the rules for expectation and variance diverge sharply. The mean of a sum is always the sum of the means: E[X + Y] = E[X] + E[Y], with no conditions — this holds whether X and Y are independent, correlated, or anything else. Linearity of expectation is unconditional and is one of the most powerful tools in probability. But the variance of a sum includes a covariance term: Var(X + Y) = Var(X) + Var(Y) + 2Cov(X, Y). Only when X and Y are uncorrelated (Cov(X, Y) = 0) does the simple additive formula Var(X + Y) = Var(X) + Var(Y) hold. Independence implies zero covariance, so independent variables always satisfy the additive formula — but the converse is not true in general.
These rules are the engine behind standardization: converting any random variable X with mean μ and standard deviation σ to Z = (X − μ)/σ, which has mean 0 and variance 1. The transformation subtracts μ (shifting the center to 0) and divides by σ (scaling the spread to 1). By the rules above, E[Z] = (E[X] − μ)/σ = 0 and Var(Z) = (1/σ)²Var(X) = σ²/σ² = 1. Standardization is the foundational step in computing z-scores, constructing confidence intervals, and working with the standard normal distribution — all of which rely on precisely these transformation rules.