σ = √(Σ(xᵢ − x̄)² / n)
The standard deviation is the square root of the variance. The work has four steps: the
mean, each value’s distance from it, the squares of those distances added up, and finally division by
n and the root of the whole. The formula is a definition.
The root is there for the unit’s sake. The variance is in squared units, and the root brings it back to
the same unit as the values. A standard deviation of 2 cm can sit next to a mean of 5 cm and be read as
“the values typically lie a couple of centimetres from the mean”.
Worked examples
Eight values. 2, 4, 4, 4, 5, 5, 7, 9 has a mean of 5. The squared distances add up to 32, the
variance is 32 / 8 = 4, and the standard deviation is √4 = 2.
With an outlier. Add 100. The mean becomes about 15.6 and the standard deviation about 29.9. One
value makes the spread fifteen times larger, because its distance from the mean is squared.
Common mistakes
- Forgetting the root. The answer is then the variance, 4, instead of 2.
- Taking the root of each term. The root is taken of the whole quotient, after the squares have been
added and divided.
- Mixing n and n − 1 without knowing which is which. The calculator’s s button and the spreadsheet’s
STDEV.S divide by n − 1 and give the sample standard deviation, 2.14
for the same list.
Common questions
What does a standard deviation of 2 mean?
That the values typically lie about 2 units from the mean, some closer and some further away. It
measures how spread out the values are, in the same unit as the values. The z-score uses
it as its yardstick.
Can the standard deviation be 0?
Yes, when all the values are equal. Then no value lies anywhere but on the mean. It can never be
negative.
n or n − 1?
n when the values are the whole group you are describing. n − 1 when they are a sample from a larger
group. For eight values the difference is about 7%; for a hundred values it is about half a percent.
Why not just the mean distance?
That exists too, as the mean absolute deviation. The standard deviation won because squares are easier
to work with further than absolute values, and because it ties in with the normal distribution. The
rule of thumb that about 68% lie within one standard deviation holds for normally distributed values,
not for any list.