How to Understand Standard Deviation

The direct answer: standard deviation is a number that tells you how far, on average, the values in a data set sit from their mean. A small value means the numbers are clustered close to the average, while a large value means they are spread out. This guide explains the formula in plain terms, shows a worked example, and shows where the idea shows up in real study problems.

Standard Deviation at a Glance

What Spread Actually Means

Two classes can earn the same average test score and tell completely different stories. Suppose both classes average 75 percent. In one class, every student scored between 72 and 78. In the other, half the students scored near 50 and half near 100. The mean hides that difference. Standard deviation exposes it.

The mean tells you where the center of the data is. Standard deviation tells you how wide the cloud of points around that center is. Together they give a quick picture of a distribution. You will meet this pair in nearly every statistics topic, from describing a single sample to building confidence intervals. The NIST engineering statistics handbook has a clear reference table of spread measures if you want the formal definitions NIST Handbook.

The Formula, Step by Step

For a full population, the standard deviation sigma is:

sigma = sqrt( (1 / N) times sum over i of (x_i minus mu) squared )

For a sample drawn from a larger group, statisticians use a slightly different version:

s = sqrt( (1 / (n minus 1)) times sum over i of (x_i minus x bar) squared )

Read the sample formula left to right. First find the mean. Then subtract the mean from every value to get the deviation of each point. Square those deviations so that negative and positive gaps do not cancel. Add them up. Divide by n minus 1 (this is Bessel's correction, which gives a better estimate of the population value). Finally take the square root so the result returns to the original units.

Variance is the step just before the square root. It equals sigma squared for a population and s squared for a sample. People usually report standard deviation rather than variance because the units match the data, which makes the number easier to interpret.

A Worked Example

Take the data set 2, 4, 4, 4, 5, 5, 7, 9. There are eight values.

Step 1: find the mean. The sum is 40, so the mean is 40 / 8 = 5.

Step 2: subtract the mean from each value and square the result.

• (2 minus 5) squared = 9
• (4 minus 5) squared = 1, and there are three 4s, so 3
• (5 minus 5) squared = 0, and there are two 5s, so 0
• (7 minus 5) squared = 4
• (9 minus 5) squared = 16

The sum of squared deviations is 9 + 3 + 0 + 4 + 16 = 32.

Step 3: population variance = 32 / 8 = 4, so population standard deviation = sqrt(4) = 2.

Step 4: sample variance = 32 / 7, about 4.571, so sample standard deviation is about sqrt(4.571), about 2.14.

So most values in this set sit about 2 units away from the mean of 5. The sample version is a touch larger because the n minus 1 divisor corrects for the fact that we estimated the mean from the same data.

The Empirical Rule

When data follow a roughly bell shaped, normal distribution, the empirical rule (also called the 68 to 95 to 99.7 rule) applies:

• About 68 percent of values fall within one standard deviation of the mean.
• About 95 percent fall within two standard deviations.
• About 99.7 percent fall within three standard deviations.

This rule is a fast way to spot outliers. A value more than three standard deviations from the mean is unusual in a normal data set. In practice, many real world data sets are not perfectly normal, so treat the percentages as approximations rather than exact counts. Heights, test scores, and measurement errors often come close enough for a quick check, but income and reaction times usually do not.

When the data is not normal

The empirical rule assumes a bell shape, and real data often is not. Income is right skewed, with a long tail of very high earners, so most people sit below the mean and the 68 to 95 to 99.7 bands do not hold. In those cases, report the median and interquartile range alongside the standard deviation, or draw a histogram, so the shape is visible instead of hidden behind one number.

Standard Deviation Versus Other Measures

Standard deviation is not the only way to describe spread. The range (max minus min) is simple but sensitive to a single extreme value. The interquartile range looks at the middle 50 percent of the data and ignores outliers entirely. Standard deviation uses every value, which makes it sensitive to extremes but also informative when the data are roughly symmetric.

Mean absolute deviation is a close cousin. Instead of squaring the deviations, it takes their absolute value before averaging. It is easier to explain but harder to use in later formulas, which is why standard deviation became the standard tool in statistics.

A Second Example With an Outlier

Take the set 10, 12, 12, 13, 14, 15, 16, 90. The mean is (10 + 12 + 12 + 13 + 14 + 15 + 16 + 90) / 8 = 182 / 8 = 22.75. So the mean sits well above most values because of the single 90.

Compute the population standard deviation. The squared deviations from 22.75 are roughly 163, 116, 116, 96, 77, 61, 46, and 4532. The sum is about 5147, divided by 8 gives variance about 643, and the standard deviation is about 25.4. That single outlier of 90 pulls both the mean and the standard deviation upward.

This is why standard deviation and the mean are reported together. If you only heard "average 22.75," you would picture a typical value near 23, but seven of eight values are below 16. The large standard deviation is the warning that the mean is not telling the whole story. In real data, checking both numbers before trusting the average is a habit worth building.

Common Misconceptions

• Forgetting to square the deviations, which lets positives and negatives cancel.
• Using N instead of n minus 1 for sample data, which understates the spread.
• Reporting variance when the question asks for standard deviation, or vice versa.
• Treating the empirical rule as exact for non normal data.
• Confusing standard deviation (spread) with standard error (spread of a sample mean).
• Thinking a bigger standard deviation is always bad. It depends on what you are measuring.

Frequently Asked Questions

Is standard deviation the same as average distance from the mean?

Not exactly. Standard deviation is the square root of the average of squared distances, not the plain average of distances. The plain average is called mean absolute deviation.

Why divide by n minus 1 for a sample?

Dividing by n minus 1 corrects a bias so the sample variance better estimates the true population variance. Using n would tend to underestimate the spread.

Can standard deviation be negative?

No. Because it is a square root of a sum of squares, standard deviation is always zero or positive. A value of zero means every data point equals the mean.

What is a large standard deviation?

It depends on the units and context. A standard deviation of 2 centimeters is small for building heights but large for a coin's thickness. Always compare it to the mean and the units.

Does standard deviation tell me about the shape of the data?

Only partly. It tells you about width, not symmetry or skewness. Two very different shapes can share the same standard deviation, so pair it with a graph.

What is the empirical rule?

For roughly bell shaped data, about 68 percent of values fall within one standard deviation of the mean, 95 percent within two, and 99.7 percent within three.