Standard Deviation: Measuring Spread Around the Mean

Two classes, same average

Two classes take the same quiz, scored out of $10$.

Class	Scores	Mean
A	$4, \; 5, \; 5, \; 6, \; 5$	$5$
B	$1, \; 9, \; 2, \; 8, \; 5$	$5$

The means match. The classes do not. Class A scored close together; Class B was all over the place.

The mean describes the center. Standard deviation describes how far values typically sit from that center, and it is the number that separates these two classes.

Why the deviations have to be squared

A natural first attempt is to measure each score’s distance from the mean and average those distances. For Class B:

Score	Deviation from $5$
$1$	$-4$
$9$	$+4$
$2$	$-3$
$8$	$+3$
$5$	$0$
Sum	$0$

The deviations add to zero, and not by coincidence. The mean is the balance point of the data, so the amounts above it exactly cancel the amounts below it. That happens for every data set there is.

Something has to stop the canceling. Squaring does it: a negative deviation squared is positive.

$$(-4)^2 = 16 \qquad (+4)^2 = 16$$

Squaring is what keeps below-the-mean and above-the-mean from erasing each other. It also weights big deviations more heavily than small ones, which is part of why standard deviation reacts so strongly to outliers.

Mean absolute deviation solves the same problem with absolute values instead. Standard deviation uses squares because they behave far better in the algebra that later statistics is built on.

Calculating it

$$\sigma = \sqrt{\frac{\sum (x - \mu)^2}{n}}$$

In words, five steps:

1. Find the mean.
2. Subtract the mean from each value.
3. Square each of those deviations.
4. Average the squares. That number is the variance.
5. Take the square root.

For Class B, with mean $5$:

Score	Deviation	Squared
$1$	$-4$	$16$
$9$	$4$	$16$
$2$	$-3$	$9$
$8$	$3$	$9$
$5$	$0$	$0$
	Sum	$50$

$$\text{variance} = \frac{50}{5} = 10 \qquad \sigma = \sqrt{10} \approx 3.16$$

For Class A the squared deviations are $1, 0, 0, 1, 0$, summing to $2$:

$$\text{variance} = \frac{2}{5} = 0.4 \qquad \sigma = \sqrt{0.4} \approx 0.63$$

Class B’s standard deviation is five times Class A’s. That matches what the lists already showed, now as a single number.

Why the square root comes last

Squaring changed the units. If the data is in inches, the squared deviations are in square inches, and so is the variance.

A spread measured in square inches cannot be compared with the data itself. The square root converts it back to inches.

That is why variance and standard deviation both exist. Variance is the useful quantity for the arithmetic; standard deviation is the one you can read alongside the data.

Population or sample

Dividing by $n$ gives the standard deviation of the whole group you measured — a population standard deviation, written $\sigma$.

When the data is a sample used to estimate a larger population, statisticians divide by $n - 1$ instead and write $s$:

$$s = \sqrt{\frac{\sum (x - \bar{x})^2}{n - 1}}$$

A sample tends to understate how spread out the population is, and dividing by a slightly smaller number corrects for that. Calculators and spreadsheets offer both, often labeled $\sigma_x$ and $s_x$. For large data sets the two are almost identical.

Choosing the right summary

Standard deviation is paired with the mean, and both share a weakness: a single extreme value pulls them hard.

Eight home prices on one street, in thousands of dollars: $240, 250, 255, 260, 270, 275, 280, 290$. Then a mansion sells for $1{,}500$.

Summary	Eight houses	With the mansion
mean	$265$	about $402$
median	$265$	$270$
standard deviation	about $16$	about $388$
IQR	$25$	$32.5$

The mean jumped by $137$ and the standard deviation grew about twenty-fourfold. The median moved $5$ and the IQR moved $7.5$.

Shape of the data	Report
roughly symmetric, no outliers	mean and standard deviation
skewed, or with outliers	median and interquartile range

Match the summary to the shape. Describing that street with a mean of $402,000 and a standard deviation of $388,000 describes houses that are not there — eight ordinary homes and one outlier. The median and IQR describe the typical house honestly.

Worked examples

Common mistakes

Practice problems