1. Standard deviation describes spread
Standard deviation (SD) measures the dispersion of observations around their mean. For data 2, 4, 4, 4, 5, 5, 7, 9, the mean is 5. Their squared deviations sum to 32. Population SD would be √(32/8)=2; sample SD uses the n−1 denominator and equals √(32/7)≈2.13809. Choose the population method only when the values represent the full population being described.
A sample SD is not divided by the number of observations a second time.
2. Standard error describes uncertainty in the sample mean
The standard error of the mean (SE or SEM) is estimated by sample SD / √n. With n=8 and sample SD≈2.13809, SE≈2.13809/√8≈0.75593. Under independent, identically distributed sampling with finite variance, a larger sample generally makes this estimate smaller. Doubling n does not halve SE: it changes by a factor of about 1/√2 if the underlying variability stays similar.
3. Which should your report show?
When describing individual scores, heights, prices or other raw measurements, SD is generally relevant. When communicating how uncertain the sample mean is as an estimate, SE is relevant. A graph with ±1 SE does not automatically represent the 68% of observations or a 95% confidence interval. Confidence intervals require an appropriate critical value, usually a t distribution for a mean with unknown population variance.
4. Assumptions matter more than the decimal places
Neither SD nor the simple SE formula handles strongly dependent observations, repeated measurements of the same person, selection bias or unusual cluster sampling. Outliers may also alter the summaries. Our Standard Deviation Calculator lets you choose population versus sample, and the Standard Error Calculator deliberately uses sample SD. For a comparison of two independent success rates, use the Two-Proportion Z Test only when its expected-count checks pass.
Frequently asked questions
Should standard error always be smaller than standard deviation?
For n greater than one, the conventional SE of the mean is sample SD divided by √n, so it is no larger than sample SD.
Does a small SE mean observations barely vary?
No. A large sample can give a small SE even if the individual observations vary widely.
Can I calculate a confidence interval as mean ± 1.96 SE?
That is a normal approximation, not a universal rule. Small samples with unknown variance often need a t critical value; nonindependent data may require another method.
Calculations run in your browser. This guide is general educational information, not individualized financial, medical or legal advice.