The standard error is a statistical quantity that describes how much a sample-based statistic (commonly the sample mean) is expected to vary from one sample to another. More precisely, it is the standard deviation of the sampling distribution of that statistic. Practitioners often use the phrase standard error both for the theoretical value (when population parameters are known) and for an estimate computed from observed data.
What it measures
Where the ordinary standard deviation summarizes spread among individual observations, the standard error summarizes spread among estimates computed from independent samples. For example, if many independent samples of size n are drawn from the same population, the sample means will form a distribution whose standard deviation is the standard error of the mean. This quantity answers the question: how precise is the sample statistic as an estimator of the corresponding population parameter?
Common formulas and estimation
When the population standard deviation σ is known, the standard error of the sample mean is σ/√n. In practice σ is usually unknown, and the sample standard deviation s is substituted, giving the estimated standard error s/√n. For other statistics there are analogous formulas. For a proportion p estimated from n observations the estimated standard error is √(p(1−p)/n). Researchers often report these estimated standard errors alongside point estimates.
- Mean: estimated SE ≈ s / √n
- Proportion: estimated SE ≈ √(p(1−p)/n)
- Regression coefficients: SEs are obtained from the model’s estimated variance–covariance matrix
Interpretation and use in inference
The standard error is central to constructing confidence intervals and performing hypothesis tests. A confidence interval around an estimated mean is typically written as estimate ± (critical value) × SE. For large samples, the sampling distribution of many statistics is approximately normal, so z-scores or normal critical values are used. When the population variance is unknown and sample sizes are small, the Student’s t distribution is appropriate and the estimated SE is combined with t critical values.
Estimation methods and modern practice
Besides analytic formulas, resampling methods such as the bootstrap can estimate standard errors without strong distributional assumptions: repeated resampling of the observed data produces an empirical sampling distribution whose standard deviation gives a bootstrap SE. Software packages routinely produce SEs along with parameter estimates, and these reported SEs are used to assess precision and statistical significance.
History and notable distinctions
The concept of variability in estimates developed alongside early work on sampling and experimental design in the 19th and 20th centuries. Key contributors to the formal foundations of sampling distributions and inference include statisticians such as Fisher and Gosset, who clarified how sample-based estimators behave. It is important to distinguish the standard error from the population standard deviation and from related notions such as standard deviation of residuals or root mean square error; each quantity describes a different aspect of variability.
Practical notes and common misconceptions
Increasing sample size reduces the standard error at the rate of 1/√n, so larger samples yield more precise estimates, all else equal. A small SE does not guarantee absence of bias: an estimator can be precise (small SE) yet systematically off target. When reading reports, note whether the SE is theoretical (uses a known σ) or estimated from data, and whether methods for correlated or clustered data were used — naive SE formulas assume independent observations.
For background on standard deviation as a related concept, see standard deviation. For an accessible discussion of averages and sample means, see average (mean).
