Variance first, square root second
Here a complete finite probability distribution is specified, with nonnegative weights summing to one.[1]
For n equally weighted values in a complete finite population divide by n. Here n≥1.[1]
A variable measured in m has variance in m2 and standard deviation in m. σ is nonnegative and is zero exactly when all values with positive probability coincide. Zero-weight cases may have arbitrary finite values without affecting the distribution. An infinite distribution needs a finite second moment for finite variance.[1][2]
When estimating an unknown population variance from n≥2 independent identically distributed observations, the usual unbiased sample variance uses denominator n−1. Its square root is the usual sample standard deviation, but is generally not itself an unbiased estimator of σ. That is a different objective from describing the fully specified values here.
Why square and center?
Ordinary signed deviations cancel. Their squares measure spread without sign cancellation.[1]
The first identity follows by expansion and E(X)=μ. The pairwise form follows from the double sum and Σpi=1.[2]
Subtracting two nearly equal large rounded numbers in E(X2)−μ2 can be numerically inaccurate. The calculator therefore combines supplied values and weights rationally before rounding the variance or root. List mode also preserves exactly typed decimal values through centering. A tiny variance must not underflow to zero before taking its root when σ itself is still representable.
Translation does not change spread; scaling acts through the absolute value. Sums of dependent variables additionally require covariance.[2]
The familiar 68–95 rule is a property of a normal distribution, not every distribution sharing μ and σ. General bounds such as Chebyshev’s inequality are a different matter; σ alone does not determine exact central percentages. The existing bell-curve illustration is therefore explicitly labelled as a normal example.
You can describe distributions with quartiles and boxplots when comparing spread without squared deviations.
A distribution and a complete list
Three model delivery times
- Given
- X is 1, 3 or 5 days with probabilities 0.25, 0.5, 0.25.
- Find
- Calculate the mean, variance and standard deviation.
First form the weighted center, then weighted squares and their square root.
Find the weighted mean.
Square and weight deviations.
Take the root only at the end.
Result: The standard deviation is about 1.414 days.
Check: . Subtracting μ2=9 again gives 2. Only value 3 lies within μ±σ here, giving 50% rather than the normal rule’s approximately 68%.
Four fully described measurements
- Given
- A complete finite population consists of 2, 4, 4 and 6 mm.
- Find
- Calculate population standard deviation and compare denominator n−1.
Center all n complete equally weighted entries using population divisor n.
Calculate the mean of the four entries.
Form the squared-deviation sum and divide by n.
Distinguish the root and the alternative sample quantity.
Result: For the specified complete population σ≈1.414 mm is correct; s answers a different estimation question.
Check: As a weighted distribution, 2 and 6 mm each have probability 1/4 and 4 mm has probability 1/2. This exactly reproduces the first example on the mm scale. Shifting all values by 100 mm would leave both spreads unchanged.
Recall the method later with the formula’s flashcards.
10 flashcards · English materials
Already have an account? Sign in for flashcardsAvailable from Essential. A ready-made formula set, no AI credits required. Article exercises are not all included.
Five exercises on spread and units
Calculate · Foundation
X takes 0 and 4 with probability 1/2 each. Find μ and σ.
Show hint
Both values are equally far from the mean.
Learning goal: Weight a two-point case.
Show solution and calculation
μ=2.
Variance=(0−2)2/2+(4−2)2/2=4, so σ=2.
Check: E(X2)=8 and μ2=4 also give variance 4.
Rearrange · Application
σX=3 cm and Y=−2 X+7 cm. What is the standard deviation of Y?
Show hint
The constant term shifts only location.
Learning goal: Apply scaling with absolute value.
Show solution and calculation
σY=|−2|·3=6 cm.
Variance is multiplied by (−2)2=4 and becomes 36 cm2.
The constant 7 cm contributes no spread.
Error analysis · Application
x1=2, x2=100, x3=−50 and p1=1, p2=p3=0. Someone claims σ is large.
Show hint
Which values can actually occur?
Learning goal: Do not count zero-weight cases as spread.
Show solution and calculation
X=2 with certainty; μ=2 and every weighted squared term is zero.
Thus σ=0.
The other listed values have zero weight and produce no spread.
Interpret · Foundation
A variance is 9 s2. What standard deviation corresponds to it?
Show hint
Take the nonnegative square root of the unit as well.
Learning goal: Distinguish variance units from standard deviation units.
Show solution and calculation
σ=√(9 s2)=3 s, not 3 s2 or ±3 s.
Standard deviation is nonnegative by definition.
Transfer · Challenge
X has variance 4 and Y=−X. What is Var(X+Y), and why is it not 8?
Show hint
The sum is zero for every outcome.
Learning goal: Do not add variances without checking dependence.
Show solution and calculation
X+Y=0 with certainty, so Var(X+Y)=0.
X and Y are perfectly negatively dependent and their covariance is −4.
The general rule 4+4+2(−4)=0 works; the independent-sum rule omitting covariance would be wrong.
Check your understanding with the formula’s reference test.
8 test questions · English materials
Already have an account? Sign in for the testAvailable from Essential. A ready-made formula set, no AI credits required. Article exercises are not all included.
Which spread is calculated here?
Expected value supplies the center. Sample variance treats the n−1 estimator for an unknown population. The Normal distribution explains the special bell-curve percentages. These three questions remain different even when the same numeric values are entered.
The three-case mode uses p3=1−p1−p2. The two list modes calculate population variance or population standard deviation for 1 to 100 equally weighted entries. Inputs must already use the same unit; interpret σ in that unit and variance in its square. No mode promises a statistically unbiased estimator of σ. Three-case mode explicitly rejects the rounding boundary where p1+p2 rounds to one but a nonzero internal remainder remains. It does not silently set that remainder to zero; this is a calculation limit rather than a mathematical error in, for example, 0.7+0.3=1. Use “Standard deviation of a complete value list” or “Variance of a complete value list” for a fully specified equally weighted population.
Sources for this explanation
Quanta · Updated November 5, 2026
- OpenStax Introductory Statistics 2e §4.2: probability-weighted expectation and standard deviation
- Harvard Stat110 midterm review, page4: linearity, LOTUS and variance with dependence
- OpenStax Introductory Statistics 2e §4.3: binomial mean and variance
Editorial responsibility: Quanta Study editorial team
Review scope
This explanation, 2 worked examples and 5 exercises are available in German and English. Structure, source references and mathematical notation undergo automated checks. Calculations, model assumptions and both language versions receive a separate AI review.
An independent, named human subject review has not yet been documented. Our exercise numbers are worked examples, rather than measurements published by the cited sources.
Editorial responsibility, sources and methodsChange history
- Added contextual links to relevant next learning steps.
- Authored a full German-English article with an original explanation, two worked examples and five varied exercises.
Cite this reference page
Quanta. Standard Deviation. Updated November 5, 2026. https://quanta-study.de/en/formula-sheet/standardabweichung
When citing a specific scientific statement, also cite the specialist source linked to that statement.