The expected value of the squared deviation of a random variable from its mean; a measure of dispersion with units equal to the square of the variable's units.

Principle

Variance = E[(X - E[X])^2]; it decomposes via the law of total variance and is additive for independent variables (Var(X+Y)=Var(X)+Var(Y) when independent).

Demonstration

For a fair six-sided die, X in {1,...,6}, E[X]=3.5 and Var(X)=E[X^2]-(E[X])^2 = 35/12 ≈ 2.9167. For a Bernoulli(p) variable, Var(X)=p(1-p).

Misapplication

Treating variance as directly comparable to typical deviation without accounting for squared units, or using sample variance formulas incorrectly (confusing biased and unbiased estimators), or applying additivity when variables are dependent.

Consequence

Variance quantifies spread in a way that interacts well with linear operations and central limit approximations; it underlies confidence intervals, hypothesis tests, and risk measures but requires caution with non-Gaussian tails.

Reversal

Consider precision (the inverse of variance) as a measure of concentration, or use mean absolute deviation as an alternative that measures typical deviation without squaring.

Boundary

Defined when the second moment exists (E[X^2]<∞). For heavy-tailed distributions with infinite second moment, variance is undefined or infinite; variance also ignores direction of deviation and is sensitive to outliers.

Semantic Tension

Variance is a mathematically convenient dispersion measure (quadratic) but can conflict with intuitive notions of 'typical error'—standard deviation or median absolute deviation may better reflect typical magnitude in some contexts.

Synthesis

Variance is the quadratic measure of spread obtained by averaging squared deviations; it is algebraically tractable and central to linear and probabilistic calculations, but its interpretation requires attention to units and moment existence.