Standard deviation describes how spread out individual observations are; standard error describes how uncertain a sample-based estimate is. In short, standard deviation is about variation in the data, while standard error is about the precision of a statistic such as a sample mean.
What Standard Deviation Measures
Standard deviation (SD) summarizes how far observations tend to lie from their mean. A small SD means values cluster relatively closely around the average; a large SD means they are more dispersed. It is expressed in the same units as the measurements, which makes it directly interpretable alongside the data.
For example, if a class records the time students take to complete a task, the mean gives the average time and the SD indicates how much students’ times typically vary around that average. The SD is not the range: it does not simply subtract the smallest value from the largest, and it is not a statement that every observation lies within one SD of the mean.
For a population with values x and population mean μ, the population SD is the square root of the average squared deviation from μ. In a sample, the usual sample SD uses deviations from the sample mean and divides the sum of squared deviations by n − 1 before taking the square root. Here, n is the sample size. The n − 1 adjustment is commonly called Bessel’s correction; it helps make the sample variance an unbiased estimator of population variance under standard sampling assumptions.
What Standard Error Measures
Standard error (SE) measures the sampling variability of an estimate. Imagine repeatedly drawing samples from the same population and calculating the same statistic each time. Those estimates would differ from sample to sample; their distribution is called a sampling distribution, and its SD is the statistic’s standard error.
In practice, researchers usually have only one sample, so they estimate the SE from that sample and an appropriate statistical model. For a sample mean from independent observations with a common population SD, the estimated SE is the sample SD divided by the square root of n. It is expressed in the same units as the mean, but it describes uncertainty in the estimated mean—not the spread of individual observations.
Standard errors also apply to other estimates, including proportions, regression coefficients, and differences between group means. Their formulas depend on the statistic, study design, and assumptions. The familiar SD divided by the square root of sample size is specifically the usual formula for a mean under suitable conditions, not a universal formula for every SE.
Standard Deviation and Standard Error Side by Side
| Feature | Standard deviation | Standard error |
|---|---|---|
| Main question | How much do observations vary? | How much would the estimate vary across samples? |
| Describes | Spread among individual data values | Precision or sampling uncertainty of a statistic |
| Common example | Spread of students’ task completion times | Uncertainty in the estimated class mean time |
| Units | Same units as the measured variable | Same units as the estimate |
| Effect of a larger sample | Not necessarily smaller; it reflects population variability | Usually smaller for a mean when added observations provide independent information |
| Common use | Describe a dataset or population | Build confidence intervals and conduct statistical tests |
The distinction is about what is being described, not merely which number is larger. An SE calculated as SD divided by the square root of n will usually be smaller than the SD when n exceeds one, but that numerical relationship does not make the two quantities interchangeable.
Concrete Example: Reading a Standard Deviation
Suppose a sample of students reports how many minutes they spend getting to campus. The sample mean is the estimated average commute time, and the sample SD summarizes how much the reported times differ from that mean. If the SD is relatively large, commutes in the sample are diverse; if it is relatively small, they are more alike.
This interpretation concerns the students’ individual reports. It does not tell us how precisely the sample mean represents all students at the institution. A dataset can have high individual variation but still estimate a mean precisely if the sample is large and appropriately selected. Conversely, a small sample can leave the mean uncertain even when the observed values appear fairly close together.
Concrete Example: Reading a Standard Error
Now consider the same sample mean. Its SE estimates how much the sample mean would vary if comparable samples were repeatedly selected and the mean recalculated. Under the usual independent-observation formula, increasing the sample size reduces the estimated SE in proportion to the square root of the sample size, if the SD remains about the same.
That reduction concerns precision, not the underlying differences among students’ commutes. Collecting more observations does not necessarily make individual commute times less variable; it generally gives a more stable estimate of their average. A confidence interval for the mean often uses the estimated SE together with a critical value from an appropriate distribution. The interval’s interpretation depends on the method and assumptions, and the SE alone is not the interval.
Why Sample Size Changes Standard Error
The square-root relationship follows from how independent observations contribute information about a mean. When observations are independent and drawn under suitable conditions, averaging more of them tends to smooth out random sample-to-sample fluctuations. If the sample size is multiplied by a factor, the SE is divided by the square root of that factor, assuming the population variability is unchanged.
This does not mean that every larger sample automatically produces a trustworthy estimate. A very large sample selected in a biased way can estimate the wrong population quantity with impressive-looking precision. Likewise, observations that are correlated—such as repeated measurements on the same person or students grouped in classrooms—do not provide as much independent information as the same number of unrelated observations. The design may require a different SE calculation.
How to Choose Which Measure to Report
Choose the measure that matches the question. If the goal is to describe how varied the observed data are, report the SD. If the goal is to communicate uncertainty around an estimated parameter or statistic, report its SE, often alongside a confidence interval. In many research reports, both are useful because they answer different questions.
- Use SD to summarize the spread of individual observations in a sample.
- Use SE to describe the estimated sampling uncertainty of a mean or another statistic.
- For a confidence interval, use the SE and the appropriate critical value rather than treating the SE itself as the interval.
- Check the study design and method used to calculate the SE, especially for clustered, weighted, or repeated-measure data.
- Label the quantity clearly so readers can tell whether a reported value is an SD or an SE.
When describing a sample, a mean with an SD communicates both its center and the observed variability. When making an inference about a population mean, a mean with a confidence interval is often more informative than a mean with an SE alone, because the interval makes the uncertainty easier to see.
Common Mistakes and Misconceptions
A frequent error is saying that the SE describes how spread out the data are. It does not: that is the SD’s role. Another is interpreting a small SE as evidence that observations are similar. A small SE may instead result from having many observations, even when individual values vary widely.
It is also incorrect to assume that a mean plus or minus one SE automatically contains a specified proportion of individual observations or provides a standard confidence interval. Those interpretations depend on the distribution, sample size, and inferential procedure. A confidence interval generally uses a multiplier, such as a critical value, rather than simply taking one SE on either side of the estimate.
Finally, students sometimes compare SEs from studies without checking whether the estimates and designs are comparable. Different sampling methods, dependence among observations, or different definitions of the target statistic can change the calculation. Always consult the methods section or the statistical software’s documentation to see what uncertainty measure was reported.
How to Remember the Difference
Think “standard deviation: data spread” and “standard error: estimate’s uncertainty.” The word error here does not mean that a researcher made a mistake or that an observation is wrong. It refers to the expected sampling fluctuation of an estimate under a statistical model.
A useful mental picture is a set of measurements and the mean calculated from them. SD describes how scattered the measurements are around their mean. SE describes how scattered the means would be across repeated samples. One concerns individual values; the other concerns a statistic derived from those values.
Frequently Asked Questions
Is standard error the same as standard deviation?
No. SD measures variation among observations; SE estimates variation in a sample statistic across repeated samples. For a sample mean under common assumptions, the estimated SE is the sample SD divided by the square root of the sample size.
Should I report standard deviation or standard error?
Report SD when you want to describe the spread of the observations. Report SE when you want to describe the sampling uncertainty of an estimate. For inference, a confidence interval is often helpful alongside the estimate.
Does a smaller standard error mean the data are less variable?
No. A smaller SE means the estimate is more precise under the calculation’s assumptions. The observations can remain highly variable; a larger, suitable sample can still estimate their mean more precisely.
Can standard error be used to make a confidence interval?
Yes. A confidence interval is commonly calculated using an estimate, its SE, and an appropriate critical value. The exact procedure depends on the statistic, sample size, and assumptions, so one SE alone is not usually the full interval.
Why does standard error decrease as sample size grows?
For a mean based on independent observations, averaging more observations generally reduces random sampling fluctuation. The usual estimated SE is the sample SD divided by the square root of the sample size, provided the assumptions behind that formula are reasonable.