Same Mean, Different Spread: Variance and Standard Deviation
Compare two datasets with the same mean, distinguish population and sample variance, and check the units of standard deviation.
Does the same mean imply the same spread?
No. A mean describes the centre; spread describes how far values lie from that centre. Before calculating, establish whether the values represent an entire population or a sample used to estimate a larger population.
Treat two datasets as complete populations
Let A contain 1, 4, 7 and B contain -2, 4, 10. Both means are 4.
| Quantity | A | B |
|---|---|---|
| Deviations from the mean | -3, 0, 3 | -6, 0, 6 |
| Sum of squared deviations | 18 | 72 |
| Population variance: divide by 3 | 6 | 24 |
| Population standard deviation |
Every deviation in B is twice the corresponding deviation in A. Squaring produces a factor of four, so the variance is four times as large while the standard deviation is twice as large. Those are different statements.
Why might another formula divide by n-1?
Population variance is commonly written . A common sample variance used for population estimation is , with more than one sample value. Under that convention, the sample variances in the example would be 9 and 36.
Different denominators represent different statistical roles, so choose the convention that fits the question.
Use units as a check
If the original observations are lengths in cm, deviations are in cm, variance is in and standard deviation returns to cm. Variance is not directly an ordinary distance from the mean.
Adding 5 to every observation raises the mean by 5 but leaves every deviation unchanged, so variance stays the same. Multiplying all observations by 2 doubles every deviation and multiplies variance by 4. Both facts follow from the definition rather than requiring separate rules without explanation.
Questions to ask before and after calculating
Beforehand: what role do the data have, and which denominator is intended? Afterwards: are the squared deviations nonnegative, are the units correct, and does the result respond sensibly to a change of scale? Equal observations must have variance zero. A negative variance signals a calculation or formula error, not a direction of spread.