Standard Deviation Calculator

Standard deviation tells how far values in a dataset typically lie from their mean, expressed in the same unit as the original measurements. This calculator handles the repeated deviation, squaring, averaging, and square-root steps used to summarize exam scores, measurements, production results, experimental observations, and other numeric collections. A small value indicates clustering around the mean; a large value indicates a wider spread.

Quick Answer

Population standard deviation is σ = √(Σ(xᵢ − μ)² / N). It is the square root of the population variance.

From Squared Spread Back to the Original Unit

The calculation begins by finding the mean, then measuring each value's deviation from that center. Squaring the deviations prevents positive and negative differences from canceling and emphasizes points that lie far away. Averaging the squared deviations produces variance. Taking the square root returns the summary to the original unit, which is the principal reason standard deviation is reported so often: a standard deviation of 4 points can be compared directly with test scores measured in points.

Standard deviation is not the average absolute distance from the mean, and it is not the range. The range uses only the smallest and largest values. Standard deviation uses every observation, although extreme values exert strong influence because their deviations are squared. It also does not say that every value lies within one standard deviation of the mean. For a roughly bell-shaped distribution, about two thirds often do, but skewed or unusual distributions can behave differently.

The denominator depends on the purpose. Use N for a complete population. Use n − 1 for the usual sample standard deviation, which estimates the spread of a larger population from a sample. The difference can be substantial in a small dataset. The Variance Calculator explains the squared measure and its denominator choice in more detail.

How to Use the Standard Deviation Calculator

1. Gather all observations for the dataset in one unit.
2. Remove duplicates only if they are data-entry errors, not merely repeated measurements.
3. Enter each value once.
4. Select or apply the population calculation when the dataset is the entire group of interest.
5. Use the sample denominator when estimating a larger population from a subset.
6. Report the standard deviation beside the mean, dataset size, and unit.

Five Test Scores Around a Mean of 84

Consider scores of 72, 78, 84, 88, and 98. Their sum is 420, so the mean is 420 / 5 = 84. Deviations from the mean are −12, −6, 0, 4, and 14. Their squares are 144, 36, 0, 16, and 196, totaling 392. Treating the five scores as the full population gives variance 392 / 5 = 78.4. The standard deviation is √78.4 = 8.85 points, rounded to two decimal places.

If these scores were instead a sample from a much larger class, the sample variance would be 392 / 4 = 98, and the sample standard deviation would be about 9.90 points. The observations did not change, but the statistical claim did: one value describes only the five scores, while the other estimates the larger group's spread.

• The dataset 10, 10, 10, and 10 has a standard deviation of 0.
• The population dataset 0 and 10 has a standard deviation of 5.
• The population dataset 9, 10, and 11 has a standard deviation of about 0.82.

Population, Sample, and Outlier Decisions

Choosing the wrong denominator is not a cosmetic issue. In a sample of five, dividing by five instead of four understates the estimated variance and standard deviation. In a complete roster, using n − 1 overstates the spread of the group actually measured. Words such as "sample," "survey respondents," "all employees," or "the measurements collected" help determine the intended target, but the analyst must decide what population the result is supposed to describe.

Outliers require investigation. A score of 98 in the example raises the result, but it is a plausible score and belongs in the summary. A recorded score of 980 caused by an entry error would overwhelm the calculation and should be corrected. Standard deviation should never be used as a reason to erase an inconvenient but genuine observation; instead, report that the data contain an extreme value and consider a median or interquartile range as an additional summary.

Do not attach a percentage sign unless the value has deliberately been converted into a relative measure. A standard deviation of 8.85 points is not "8.85%." Also avoid comparing standard deviations across datasets with different means or units without context. To standardize an individual score against a mean and spread, continue with the Z-Score Calculator. More summaries are available in Mathematics Calculators.

Why the Mean Must Travel With the Spread

A standard deviation of 5 has different practical meaning for measurements centered at 20 than for measurements centered at 500. In the first case, typical squared-deviation distance is large relative to the center; in the second, it is comparatively modest. This is why analysts sometimes calculate a coefficient of variation by dividing standard deviation by the mean, but that relative measure is meaningful only for positive ratio-scale data and becomes unstable when the mean is near zero. It is a supplement, not a replacement for the original units.

Comparing two groups requires the same care. A training program might reduce average completion time while also reducing its standard deviation, suggesting both faster and more consistent performance. Another might lower the mean by helping only a few participants dramatically, increasing spread. Standard deviation exposes that distinction, but it does not establish why it happened or whether the difference matters operationally. Pair the summary with group sizes, context, and the underlying observations when a decision follows.

For repeated measurements, calculate the standard deviation separately for each instrument, operator, or condition before pooling the data. Combining groups with different centers can create a large overall spread even when each group is internally consistent. Separate summaries make that structure visible and point to calibration or process differences. Pool only when the groups are meant to represent the same underlying population and measurement method.

A Practical Way to State the Result

A complete report sounds like this: "The five scores had a mean of 84 points and a population standard deviation of 8.85 points." The center, spread, count, unit, and denominator choice are all present. That context matters more than an extra decimal place.