Skip to content

Variance and Standard Deviation

Definition / Introduction

  • While Expected Value \(E[X]\) describes the center of a random variable's distribution, Variance \(Var(X)\) measures its spread or dispersion.
  • It quantifies how much the values of the random variable \(X\) tend to deviate from their mean \(\mu = E[X]\) on average. A higher variance means values are more spread out; lower variance means they are more clustered around the mean.
  • Standard Deviation \(SD(X)\) or \(\sigma_X\) is simply the square root of the variance. It's often preferred for interpretation because it's in the same units as the random variable \(X\).
  • Notation: Variance \(Var(X)\), \(\sigma^2\), or \(\sigma_X^2\). Standard Deviation \(SD(X)\), \(\sigma\), or \(\sigma_X\).

Key Concepts

1. Definition based on Expected Squared Deviation

  • The variance is defined as the expected value of the squared difference between the random variable \(X\) and its mean \(\mu = E[X]\): $$ Var(X) = E[(X - \mu)^2] $$
  • Calculation (Discrete): \(Var(X) = \sum_{\text{all } x} (x - \mu)^2 p(x)\)
  • Calculation (Continuous): \(Var(X) = \int_{-\infty}^{\infty} (x - \mu)^2 f(x) \, dx\)

2. Computational Formula (Often Easier)

  • A more convenient formula for calculating variance is derived from the definition using linearity of expectation: $$ Var(X) = E[X^2] - (E[X])^2 $$ $$ Var(X) = E[X^2] - \mu^2 $$
  • This requires calculating two expected values: \(E[X]\) (the mean) and \(E[X^2]\) (the expected value of X squared, calculated using LOTUS as \(\sum x^2 p(x)\) or \(\int x^2 f(x) dx\)).

3. Standard Deviation

  • The standard deviation is the positive square root of the variance: $$ SD(X) = \sigma = \sqrt{Var(X)} = \sqrt{E[(X - \mu)^2]} $$
  • Interpretation: \(\sigma\) measures the typical or average distance of the values of \(X\) from their mean \(\mu\). A smaller \(\sigma\) means data points are typically close to the mean; a larger \(\sigma\) means they are typically far from the mean.

4. Properties of Variance

  • Non-negativity: \(Var(X) \ge 0\). Variance is zero if and only if \(X\) is a constant (no spread).
  • Constants: \(Var(b) = 0\) for any constant \(b\).
  • Linear Transformation: For constants \(a\) and \(b\): $$ Var(aX + b) = a^2 Var(X) $$ Note: Adding a constant \(b\) shifts the distribution but doesn't change its spread, so \(b\) disappears. Multiplying by \(a\) scales the deviations, and squaring \(a\) reflects the squaring in the variance definition.
    • Consequently: \(SD(aX + b) = |a| SD(X)\).

5. Variance of a Sum of Random Variables

  • For any random variables \(X\) and \(Y\): $$ Var(X + Y) = Var(X) + Var(Y) + 2 Cov(X, Y) $$ where \(Cov(X, Y)\) is the Covariance between X and Y.
  • If X and Y are Independent: Then \(Cov(X, Y) = 0\), and the formula simplifies significantly: $$ Var(X + Y) = Var(X) + Var(Y) \quad (\text{if X, Y independent}) $$ $$ Var(X - Y) = Var(X) + Var(Y) \quad (\text{if X, Y independent}) $$ Note: Variance ADDS even when subtracting independent variables because subtraction can still increase the range of outcomes.

Connections to Other Topics & Relevance

  • Risk Assessment: In finance and business, variance and standard deviation are key measures of risk or volatility (e.g., of investment returns).
  • Data Analysis & Feature Scaling: Standard deviation is used in standardization (calculating Z-scores: \(Z = (X - \mu) / \sigma\)) which is essential for many machine learning algorithms. Understanding variance helps interpret feature importance and variability.
  • Confidence Intervals & Hypothesis Testing: Standard deviation (often estimated from samples as the standard error) is crucial for constructing confidence intervals and performing hypothesis tests about population means.
  • Normal Distribution & Empirical Rule: Standard deviation defines the intervals (\(\mu \pm k\sigma\)) containing specific percentages of data (68%, 95%, 99.7%).
  • Process Control: Used to monitor the stability and consistency of processes (e.g., manufacturing).

Summary

  • Variance \(Var(X)\) measures the expected squared deviation from the mean \(E[X]\), quantifying spread. \(\sigma^2 = E[(X - \mu)^2]\).
  • Standard Deviation \(SD(X) = \sigma = \sqrt{Var(X)}\) measures the typical deviation from the mean, in the original units.
  • Calculation: \(Var(X) = E[X^2] - (E[X])^2\).
  • Properties: \(Var(aX + b) = a^2 Var(X)\). For independent RVs, \(Var(X + Y) = Var(X) + Var(Y)\).
  • Fundamental for understanding risk, variability, and for many statistical procedures.

Sources