Variance and Standard Deviation: Formulas Made Simple

Variance is the average of the squared distances of observations from the mean, and standard deviation (SD) is its square root. Together they are the standard, universal measure of how spread out data is — the SD returns the spread to the original units.

They matter because almost every advanced statistic — from z-scores and confidence intervals to regression and quality control — is built on variance. It is the single most used spread measure in science, finance and engineering.

Key Concepts

Variance (σ²): Σ(x − x̄)²/n — in squared units (e.g. marks²). Standard deviation (σ): √variance — in original units, directly comparable with the mean. Population vs sample: population variance divides by N; sample variance (s²) divides by n−1 (Bessel's correction) to remove bias. Why square: squaring stops deviations cancelling and penalises large deviations more — but makes variance hard to interpret, hence the square root.

Formulas

Population variance: σ² = Σ(xᵢ − μ)²/N. Computational form: σ² = (Σx²/N) − μ². Sample variance: s² = Σ(xᵢ − x̄)²/(n−1). Standard deviation: σ = √σ². Grouped data: σ² = Σf(x − x̄)²/Σf with class marks x. Empirical rule (normal data): ≈68% of values within x̄ ± σ, ≈95% within x̄ ± 2σ, ≈99.7% within x̄ ± 3σ. Variance units are squared; SD units match the data.

Worked Example

Marks: 42, 48, 55, 58, 61, 66, 73; mean x̄ = 57.57. Step 1 — squared deviations: (42−57.57)² = 242.4, (48−57.57)² = 91.6, (55−57.57)² = 6.6, (58−57.57)² = 0.2, (61−57.57)² = 11.8, (66−57.57)² = 71.0, (73−57.57)² = 238.0. Step 2 — sum: 242.4 + 91.6 + 6.6 + 0.2 + 11.8 + 71.0 + 238.0 = 661.6. Step 3 — variance (population): σ² = 661.6/7 ≈ 94.51 marks². Step 4 — SD: σ = √94.51 ≈ 9.72 marks. Check with computational form: Σx² = 1764+2304+3025+3364+3721+4356+5329 = 23863; 23863/7 − 57.57² = 3409.0 − 3314.3 ≈ 94.7 — matches. Typical scores fall roughly 10 marks either side of the mean.

Common Misconceptions

1. "Variance and standard deviation are the same." Variance is the squared quantity (marks²); the SD is its square root in real units — never compare a variance directly with the mean. 2. "Dividing by n−1 is just a convention." For samples it corrects a systematic underestimate; use n for a full population, n−1 for a sample. 3. "A bigger mean means a bigger SD." Spread is independent of level — a data set centred at 1000 can have a tiny SD.

Real-World Applications

Standard deviation measures investment risk (volatility), manufacturing tolerance, exam score consistency, weather variability, and drug-trial response spread. Variance powers ANOVA tests, regression models, Six Sigma quality programs and every confidence interval in research.

Variance averages squared deviations from the mean; its square root, the standard deviation, restores the original units. Divide by N for populations and n−1 for samples, and you hold the foundation of nearly all inferential statistics.