Standard Deviation Calculator
Paste a data set to calculate sample or population standard deviation, variance, mean, quartiles, z-scores, and the full calculation steps.
Calculator is for informational purposes only. Terms and Conditions
Use sample mode when the entered observations are a subset used to estimate a larger population; use population mode when they are the entire group of interest.
Choose sample or population
This choice changes the denominator in the variance and standard deviation formulas.
Enter your data
Paste values from a spreadsheet or type them directly. Results update automatically.
Separate values with commas, spaces, tabs, semicolons, or line breaks. Frequency shorthand is supported: 10:3 means the value 10 appears three times.
Result
Standard deviation is shown first, followed by variance and descriptive statistics.
Result details
Show calculation steps Review the mean, deviations, variance, standard deviation, and checks
- Enter valid data to see the complete calculation.
Per-value deviations and z-scores
| Value | Deviation | Squared deviation | Z-score | IQR check |
|---|
Data Spread Around the Mean
Each point is an observation. The center line marks the mean; adjacent reference lines mark one standard deviation below and above the mean.
Method, Sources, and Assumptions
Calculation basis, statistical conventions, limitations, and authoritative references.
Standard deviation is the square root of variance. Sample standard deviation uses n − 1 in the denominator; population standard deviation uses N. The calculator computes deviations from the mean directly and does not assume the data are normally distributed.
- Sample mode requires at least two observations; population mode permits one observation, for which the population standard deviation is zero.
- Quartiles use the median-of-halves convention used by this calculator; other software may use a different quartile interpolation convention.
- The 1.5 × IQR rule identifies potential outliers only; it does not determine whether a value is erroneous or should be removed.
- The mean ± 1 SD visual is descriptive. Standard deviation does not imply that the data follow a normal distribution.
Calculator guide
Understanding Your Standard Deviation Result
The calculator above measures how widely your data values are spread around their mean. Enter one numeric dataset, choose whether it represents a sample or the full population of interest, and the primary result is standard deviation in the same units as your data. The calculator also reports variance, mean, count, quartiles, range, IQR, z-scores, and an optional IQR outlier screen. Standard deviation is always zero or positive: zero means every observation is identical, while a larger value means greater spread on that dataset’s scale.
Standard deviation is useful because it converts many individual deviations from the mean into one summary of variability. NIST describes standard deviation as the square root of variance and notes that it restores the measure of spread to the original data units, whereas variance uses squared units. NIST’s Measures of Scale provides the statistical basis for that interpretation.
- Input
- A numeric dataset, including decimals, negative values, scientific notation, or value:frequency shorthand.
- Primary output
- Sample standard deviation \(s\) or population standard deviation \(\sigma\).
- Key decision
- Use sample mode for a sample used to represent a larger population; use population mode for the complete population you intend to describe.
Sample vs. Population Standard Deviation
The calculation changes only in the variance denominator, but that choice matters most for small datasets. In sample mode, the calculator divides the sum of squared deviations by \(n-1\); in population mode, it divides by \(N\). OpenStax gives the same sample and population definitions and symbols. See OpenStax’s measures of spread.
| Question | Sample | Population |
|---|---|---|
| When to use it | Your observations are a sample used to characterize or estimate a larger population. | Your observations are the entire population you want to describe. |
| Standard deviation symbol | \(s\) | \(\sigma\) |
| Mean symbol | \(\bar{x}\) | \(\mu\) |
| Variance denominator | \(n-1\) | \(N\) |
Sample standard deviation formula
Plain language: subtract the sample mean from each observation, square those deviations, add them, divide by one less than the sample size, and take the square root.
The \(n-1\) denominator is commonly called Bessel’s correction. It makes the conventional sample variance unbiased for population variance under standard random-sampling assumptions; the square-rooted sample standard deviation itself is not generally exactly unbiased.
Population standard deviation formula
Plain language: find each observation’s deviation from the population mean, square and sum the deviations, divide by the population size, then take the square root.
- \(x_i\)
- An individual observation in the dataset.
- \(\bar{x}\)
- The arithmetic mean of a sample.
- \(\mu\)
- The arithmetic mean of the population being described.
- \(n\)
- The number of observations in the sample.
- \(N\)
- The number of observations in the population.
- \(s^2,\ \sigma^2\)
- Sample and population variance, expressed in squared data units.
How to calculate standard deviation by hand
Find the mean
Add all observations and divide by the number of observations.
Find each deviation
Subtract the mean from every observation.
Square the deviations
Square each deviation so negative and positive differences do not cancel.
Add the squared deviations
This total is often called the sum of squares.
Divide to get the variance
Divide by \(n-1\) for a sample or by \(N\) for a population.
Take the square root
The square root of variance is standard deviation, which returns the measure of spread to the original data units.
Standard deviation vs. variance
| Standard deviation | Variance |
|---|---|
| Square root of variance | Square of standard deviation |
| Same units as the original data | Squared data units |
| Usually easier to interpret directly | Often convenient in statistical formulas |
| \(s\) or \(\sigma\) | \(s^2\) or \(\sigma^2\) |
Check standard deviation in Excel
For a sample, Excel uses =STDEV.S(range). For the complete population of interest, use =STDEV.P(range). Microsoft documents that STDEV.S uses the \(n-1\) method and recommends STDEV.P when the values represent the entire population. This is a useful independent check when your data already lives in a spreadsheet.
Worked Standard Deviation Example
The calculator loads the classic dataset 2, 4, 4, 4, 5, 5, 7, 9 as an example. In the default sample mode, it returns a sample standard deviation of about 2.1381. The same observations in population mode give exactly 2.
Calculation
- Add the observations: the sum is 40, so the mean is \(\bar{x}=40/8=5\).
- Subtract 5 from each observation and square each deviation. The squared deviations are 9, 1, 1, 1, 0, 0, 4, and 16.
- Add the squared deviations: \(SS=32\).
- For sample variance, divide by \(n-1=7\): \(s^2=32/7\approx4.5714286\).
- Take the square root: \(s=\sqrt{32/7}\approx2.1380899\), displayed as 2.1381 with automatic precision.
Result
Sample standard deviation = 2.1381 data units
The root-mean-square spread of the observations around the mean of 5 is 2.1381 data units under the sample standard deviation definition.
How to Interpret Standard Deviation
Standard deviation tells you the magnitude of spread, not whether that spread is automatically acceptable. A value is meaningful only in the context of the measurement units, the mean, the distribution shape, and the engineering or statistical decision you are making.
Read it in the original units
If the dataset is in millimeters, standard deviation is in millimeters. If measurements are in MPa, standard deviation is in MPa. Variance is different: it uses squared units such as mm² or MPa².
Compare spread with context
A standard deviation of 5 cannot be labeled universally “high” or “low.” Compare it with the scale of the measurements, expected process variation, measurement resolution, and any applicable acceptance criteria.
Use the visual to inspect shape
The calculator plots observations around the mean and marks mean ± 1 standard deviation. For large datasets, the display may summarize or sample plotted observations for readability while calculations continue to use the full accepted dataset.
Normal-distribution interpretation is conditional
You can calculate standard deviation for data that are not normally distributed. The familiar 68–95–99.7 rule is a separate interpretation that applies to an approximately normal distribution: about 68% of values are within one population standard deviation of the mean, about 95% within two, and about 99.7% within three. OpenStax’s standard normal distribution section states this normal-distribution condition explicitly.
Z-scores put individual observations on the standard-deviation scale
The calculator’s per-value results can report deviations and z-scores, with the displayed table limited for very large datasets to protect performance. A z-score describes how many calculated standard deviations an observation lies above or below the mean: positive values are above the mean and negative values are below it. A large absolute z-score signals distance from the mean, but it is not by itself proof that the observation is erroneous.
Quartile conventions can change borderline IQR flags
The calculator uses a median-of-halves quartile convention for Q1 and Q3. Other statistical packages may use interpolation-based quartiles, so a value close to an IQR fence can be flagged differently between tools even when the underlying dataset is identical. Treat the IQR result as a screening aid rather than an automatic data-deletion rule.
Engineering use: variability is not the same as compliance
NIST notes that manufacturing and measurement processes exhibit variability in quantities such as dimensions, thickness, and resistivity. Standard deviation is useful for quantifying that spread, but a single SD does not establish that a process is stable or that parts meet specification limits. Process-control decisions require the time structure of the data and an appropriate control-chart or statistical method; specification compliance requires the actual specification limits. NIST’s process variability guidance explains the distinction between stable controlled variation and changing uncontrolled variation.
Common Mistakes and Important Limits
Most bad standard deviation results come from defining the dataset incorrectly rather than from the arithmetic. Check these issues before treating the output as evidence about a larger process or population.
Choosing sample or population by habit
Define the population from the question you are answering. A complete set of 12 monthly observations can be a population if your question concerns exactly those 12 months, even though future months will exist.
Mixing units or measurement conditions
Do not combine inches with millimeters, or measurements taken under materially different conditions, without first converting or segmenting the data. Standard deviation assumes the numerical observations are comparable.
Deleting a flagged outlier automatically
The calculator can flag observations outside \(Q_1-1.5IQR\) or \(Q_3+1.5IQR\). NIST uses these inner box-plot fences for identifying unusual tail values, but a flag is a reason to investigate context, not an instruction to delete the observation. See NIST’s outlier guidance.
Assuming more decimal places mean more accuracy
The calculator keeps full internal precision and lets you change displayed decimal places. Reporting more digits does not improve the quality of the measurements or the representativeness of the sample.
Treating a small sample as a precise estimate
Sample standard deviation can be computed from two or more observations, but there is no universal sample-size threshold that makes the estimate “reliable.” Sampling design, population shape, desired confidence, and the decision being made determine how much data is enough.
Confusing standard deviation with standard error
Standard deviation describes spread among observations. Standard error describes sampling variability of a statistic. For the sample mean, OpenStax gives the population-standard-deviation form \(SE_{\bar{x}}=\sigma/\sqrt{n}\); the two quantities answer different questions.
Ignoring drift or mixed operating states
If a process mean changes over time, one pooled standard deviation can combine ordinary short-term variation with drift or regime changes. Segment, detrend, or model the time structure when the purpose is to characterize stable process noise.
Assuming standard deviation proves normality
It does not. Highly skewed, multimodal, or heavy-tailed data can have a perfectly valid standard deviation while the mean-and-SD summary hides important structure. Use the calculator’s quartiles and data-spread visual as additional evidence.
Sources and Calculation Checks
The guide and calculator use standard descriptive-statistics definitions. The worked example was independently checked by recomputing the mean, squared deviations, variance, and square-root result, then reversing the variance calculation back to the sum of squared deviations.
- NIST/SEMATECH Engineering Statistics Handbook — Measures of Scale — supports the definitions and interpretation of variance, standard deviation, range, and sensitivity of variance to large deviations.
- OpenStax Introductory Statistics — Measures of the Spread of the Data — supports sample and population notation, the \(n-1\) and \(N\) formulas, frequency forms, and the distinction between standard deviation and standard error.
- NIST/SEMATECH Engineering Statistics Handbook — Outliers — supports the 1.5 × IQR inner-fence convention used by the calculator’s optional potential-outlier screen.
- NIST/SEMATECH Engineering Statistics Handbook — Process Variability — supports the engineering context for measurement and manufacturing variability and the need to distinguish stable from changing variation.
- Microsoft Support — STDEV.S function — documents Excel’s sample standard deviation function and its \(n-1\) method.
- Microsoft Support — STDEV.P function — documents Excel’s population standard deviation function for data representing the entire population.
Standard Deviation Calculator FAQ
These answers cover practical follow-up questions that are not obvious from the result alone.
Can standard deviation be negative?
No. Variance is built from squared deviations and is therefore nonnegative; standard deviation is the nonnegative square root of variance. A result of zero occurs when every observation is identical.
Can I paste frequency data into this calculator?
Yes. Use value:frequency entries such as 10:3, 12:2, 15:1. Frequencies must be positive whole numbers. The calculator expands the frequencies and includes every represented observation in the same sample or population calculation.
Why does sample mode require at least two observations?
The sample variance denominator is \(n-1\). With one observation, that denominator becomes zero, so the conventional sample variance and sample standard deviation are undefined. Population mode can describe a one-value population, whose variance and standard deviation are both zero.
How do I check the result in Excel?
For sample data, Excel uses =STDEV.S(range). For the complete population of interest, use =STDEV.P(range). Microsoft documents that STDEV.S uses the \(n-1\) method and directs users to STDEV.P when the values represent the entire population.
Does a larger standard deviation always mean worse performance?
No. Standard deviation quantifies spread; it does not define what is acceptable. Whether greater variability is undesirable depends on the variable, units, engineering objective, specification limits, process behavior, and consequences of variation.
Should I remove every value the IQR check flags?
No. The 1.5 × IQR rule is a screening tool for unusual observations. Check for data-entry mistakes, measurement problems, changed operating conditions, or a genuine rare event before deciding whether exclusion is justified, and document any exclusion rule used in formal analysis.
Why might this calculator and another program show slightly different outlier flags?
Different software can use different quartile definitions. This calculator uses a median-of-halves method for Q1 and Q3, while some programs use interpolation. Standard deviation itself can still match even when a borderline IQR outlier classification differs.