Distinguish estimation from description
Observations must share a scale and unit. The data can be summarized algebraically even without a distributional assumption.[1]
Calling it an unbiased variance estimator requires independent identically distributed observations with finite variance. Normality is not needed for this property. For dependent time series, unequal weights or different distributions, n−1 alone does not establish unbiasedness. “Corrected” refers to Bessel’s correction rather than repair of erroneous measurements.[1][2]
v describes this equally weighted finite collection. Both formulas can be meaningful in their appropriate context.[1]
If values use mm, s2 has unit mm2 and s=√s2 has unit mm. s2 is nonnegative and, for n≥2, is zero precisely when all observations coincide. For n=1 corrected variance is undefined; a single observation does not define this variance estimator.[1]
Why n minus one appears
Expansion and Σ(Xi−X̄)=0 give this decomposition, first as an algebraic identity.
For iid observations E(X̄)=μ and Var(X̄)=σ2/n. Dividing by n−1 therefore gives E(s2)=σ2.[2]
The n deviations sum to zero, so knowing n−1 determines the last. This illustrates the lost degree of freedom. The expectation calculation additionally shows precisely under which assumptions the correction removes bias. Counting degrees of freedom alone does not replace those assumptions.
Unbiasedness of s2 does not automatically transfer to s. For example, two independent observations are each equally likely to be −1 or 1. Matching observations give s=0 and differing observations give s=√2, each with probability 1/2. Thus E(s)=1/√2<1 although population standard deviation is one.
Numerically, centering is safer than subtracting two early-rounded large squared sums. The list calculator retains typed decimal values exactly through centering and rounds only the requested variance or root. The limit of 100 entries belongs to the calculator, not the mathematical formula.
A spread and a large common offset
Four observations in mm
- Given
- A sample contains 1, 3, 4 and 8 mm. The estimation interpretation assumes iid observations.
- Find
- Find corrected variance, standard deviation and descriptive variance.
First state the general setup for the supplied quantities, then substitute the particular numbers.
Form mean and deviations.
Divide the squared-deviation sum by n−1.
Calculate the root and descriptive variant.
Result: s2≈8.66667 mm2 and s≈2.94392 mm; descriptive variance of these four values is 6.5 mm2.
Check: Deviations sum to −3−1+0+4=0. The identity (n−1)s2=n·v gives 26 mm2 on both sides. Variance and its square root must not share the same unit.
Exact dimensionless decimals with a large offset
- Given
- The exact values are 10000000000000000, 10000000000000001 and 10000000000000002.
- Find
- Calculate s2 without losing the small differences.
First state the general setup for the supplied quantities, then substitute the particular numbers.
Separate the common offset symbolically.
Calculate with the small centered differences.
Divide by the two degrees of freedom.
Result: Variance and standard deviation are both numerically one; centering remains exact.
Check: Subtracting M leaves 0, 1 and 2 with the same variance. Rounding the middle typed value to a neighbouring machine value before calculation would create a different dataset. The exact-decimal list path preserves all three inputs here.
Recall the method later with the formula’s flashcards.
9 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 denominator and transformation
Calculate · Foundation
Find s2 and s for 2, 5 and 8.
Show hint
The mean is 5.
Learning goal: Center a small squared sum.
Show solution and calculation
Deviations −3, 0, 3 give Q=18.
Thus s2=18/2=9 and s=3.
Check: descriptive variance would be 18/3=6, with different denominators serving different goals.
Rearrange · Foundation
n=6 and s2=4 cm2. Find Q=Σ(xi−x̄)2.
Show hint
Multiply by n−1.
Learning goal: Recover the squared-deviation sum.
Show solution and calculation
Q=(6−1)·4=20 cm2.
Check: 20/5=4 cm2; the factor is five rather than six.
Interpret · Application
Compare data (7,7,7,7) with the single value (7).
Show hint
Check n−1 and Q separately.
Learning goal: Distinguish zero spread from absent degrees of freedom.
Show solution and calculation
Four equal values have n−1=3 and Q=0, so s2=0.
A single value has zero denominator, making corrected sample variance undefined.
Zero and undefined are different outcomes.
Error analysis · Application
Someone claims: “Dividing by n is always wrong.” Correct the statement.
Show hint
Which population is being described or estimated?
Learning goal: Name the estimation objective.
Show solution and calculation
n is correct for descriptive variance of the fully specified equally weighted collection.
The usual unbiased estimate of unknown population variance from iid data uses n−1.
Without specifying the goal, the blanket claim is false.
Transfer · Challenge
For yi=−2 xi+5, s2x=9. Find s2y and sy.
Show hint
The mean follows the same affine transformation.
Learning goal: Prove affine scaling using deviations.
Show solution and calculation
ȳ=−2 x̄+5, so yi−ȳ=−2(xi−x̄).
The squared sum is multiplied by four and n−1 is unchanged.
Thus s2y=36 and sy=6.
The shift by five does not affect spread.
Check your understanding with the formula’s reference test.
7 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.
Two list modes with explicit estimation meaning
The calculator has separate targets for s2 and s from 2 to 100 values. Repeated values are allowed; a one-value list is rejected. Algebraic calculation does not check independence, identical distribution or representativeness. A numeric result alone is therefore not a justified uncertainty estimate.
Standard deviation of a fully specified population uses its supplied weights. Expected value distinguishes model mean and sample mean. A Confidence interval for the mean uses s/√n as estimated standard error rather than s2 itself.
Sources for this explanation
Quanta · Updated November 1, 2026
- NIST Engineering Statistics Handbook §1.3.5.6: sample variance, standard deviation and units
- Harvard Stat110 midterm review page4: expectation linearity 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
- Authored a full German-English article with an original explanation, two worked examples and five varied exercises.
Cite this reference page
Quanta. Sample Variance. Updated November 1, 2026. https://quanta-study.de/en/formula-sheet/stichprobenvarianz
When citing a specific scientific statement, also cite the specialist source linked to that statement.