How this instrument works
Standard error answers a specific, practical question: if you repeated your study with a fresh sample of the same size, how much would the mean bounce around? It is not a measure of how spread out your raw data is — that job belongs to the standard deviation. Standard error is a measure of how spread out the sample mean itself would be across many hypothetical repeats of your sampling. Confusing the two is one of the most common mistakes in reported statistics: a large standard deviation just means your individual measurements vary a lot; a large standard error means you cannot trust the mean you calculated to be close to the true population mean.
The formula, SE = s ⁄ √n, ties the two ideas together directly: it takes the spread you already have (s, the sample standard deviation) and shrinks it by the square root of your sample size. Ten times the data does not give you ten times the precision — it gives you about 3.16 times the precision, because the square root grows so much slower than n itself. That is why doubling a small pilot study's sample size noticeably tightens its standard error, while doubling an already-large survey barely moves the needle.
In practice, standard error is what turns a single number into a defensible claim. A reported result of “mean = 24.3” says nothing about how much to trust it; “mean = 24.3, SE = 0.8” tells a reader the estimate is precise to within roughly a point. Scientific papers report it beside means, pollsters build margins of error from it, and error bars on charts are frequently drawn at one standard error above and below the point estimate.
- Paste or type your numbers into the Data set field — commas, spaces, or new lines all work as separators.
- The instrument first computes the sample standard deviation s, then divides it by the square root of your count n.
- Read the result in Standard error (SE) — report it alongside your mean as “x̄ ± SE” in a paper, poll write-up, or lab notebook.
- Add more data points and watch SE shrink — you need roughly four times the sample size to cut SE in half, not twice.
- Needs at least 2 values: standard error is undefined for a single observation, since there is no spread to measure yet.
Worked example — the eight quiz scores
Paste the data set {2, 4, 4, 4, 5, 5, 7, 9} — eight quiz scores, n = 8 — into Data set. The instrument first finds the sample standard deviation, s ≈ 2.13809, from the squared deviations around the mean of 5. Standard error (SE) then reads 0.755929, computed as 2.13809 ⁄ √8 = 2.13809 ⁄ 2.82843.
Report that as “mean = 5.0, SE = 0.76.” It says that if this same eight-question sample were drawn again from the same population of scores, the resulting mean would typically land within about three-quarters of a point of 5 — a much smaller number than the raw standard deviation, because SE describes the mean's precision, not the scores' spread.
Questions
What is the difference between standard deviation and standard error?
Standard deviation (s) measures how spread out your individual data points are around the mean. Standard error (SE = s ⁄ √n) measures how precise your sample mean is as an estimate of the true population mean. They answer different questions and SE is always smaller than s for any sample of 2 or more, shrinking further as n grows — s does not shrink just because you collect more data.
Why does standard error shrink as my sample size grows?
Because it is divided by √n, not n. Quadrupling your sample size only halves SE, since √4 = 2; to cut SE in half again you would need to quadruple the sample size once more. This diminishing-returns pattern is exactly why very large surveys still carry a small but nonzero margin of error — you can shrink SE, but never quite reach zero with a finite sample.
How do I turn standard error into a confidence interval?
For a reasonably large sample, an approximate 95% confidence interval is the mean ± 1.96 × SE, based on the normal approximation to the sampling distribution. For small samples a t-distribution critical value (larger than 1.96) is the more careful choice. Either way, the interval is built directly from the SE this instrument reports — it is the building block, not the finished interval.
Is standard error the same as a poll's margin of error?
Closely related but not identical. A published margin of error is typically SE multiplied by a critical value (often about 1.96 for 95% confidence), so it is slightly wider than SE alone. If a poll reports a margin of error without stating the confidence level, it is worth asking what multiplier was used before comparing it directly to a raw SE figure.
Can I compute a meaningful standard error from just 2 or 3 values?
Technically yes — the formula only requires n ≥ 2 — but the resulting SE is itself a very rough estimate when n is tiny, since s from 2 or 3 points is unstable. Treat an SE calculated from a handful of observations as a rough sense of precision, not a tight guarantee, and collect more data before relying on it for a firm claim.
Should I report standard deviation or standard error in my results?
Report standard deviation when you want to describe how varied the individual measurements are. Report standard error when you want to describe how precisely you have pinned down the mean. Many papers report both — s to describe the data, SE (or a confidence interval built from it) to describe the estimate — because they answer genuinely different reader questions.