Standard Error Calculator

Compute the standard error of a sample mean or a sample proportion — the number that tells you how precisely your sample statistic estimates the population value.

Standard Deviation vs. Standard Error at a Glance

  • Standard deviation (SD) describes how much individual observations vary from each other.
  • Standard error (SE) describes how much the sample statistic (the mean or proportion) would vary across repeated samples.
  • SE is always smaller than SD for n > 1, and it shrinks as the sample grows — SD does not.

Standard Deviation vs. Standard Error, Properly

Confusing these two statistics is probably the single most common mistake in introductory statistics, and the two are easy to untangle once you name what varies. Standard deviation answers: “if I pick one individual from this data, how far from the average is it likely to be?” It is a property of the data itself. Collecting more observations does not systematically make it smaller — it just estimates it more accurately.

Standard error answers a different question: “if I repeated this whole study with a fresh sample, how much would the computed mean move around?” It is a property of the estimate, not of individuals. Because averaging cancels out individual randomness, the mean is far more stable than any single observation, and it becomes more stable the more observations you average. That is why SE = SD/√n: same numerator, but divided by a factor that grows with sample size.

A practical consequence for reading charts: error bars showing ±1 SD describe the spread of the data, while error bars showing ±1 SE (or a confidence interval) describe the uncertainty of the mean. Papers do not always label which one they used, and the visual difference can be dramatic.

Formulas and Symbols

The calculator implements the two most frequently needed standard errors:

Standard Error Formulas:

SE of the mean = s / √n

SE of a proportion = √(p(1 − p) / n)

Symbols:

  • • s = sample standard deviation (n − 1 denominator)
  • • n = sample size
  • • p = sample proportion, as a decimal

In raw-data mode, the calculator first computes n, the mean, and the sample standard deviation for you, then divides by √n. It deliberately uses the n − 1 (Bessel-corrected) standard deviation, because when data is a sample used to infer something about a population, the corrected estimate is the appropriate input to the SE formula.

Why √n Sits in the Denominator

When independent measurements are averaged, their random errors partially cancel: some observations land above the truth, others below. Probability theory quantifies the cancellation — the variance of a mean of n independent values is the individual variance divided by n, so its standard deviation is divided by √n. The square root is the reason precision is expensive: each halving of the standard error demands four times the data.

Concretely, a sample with s = 15 gives SE = 15/√36 = 2.5 at n = 36, but you need n = 144 to reach SE = 1.25. Diminishing returns like this are exactly why study planning starts from a target precision rather than from a convenient sample size.

Where the Standard Error Goes Next

The standard error is rarely the final answer; it is the building block of inference. Multiplying it by a critical value gives the margin of error, and attaching that margin to the estimate gives a confidence interval. Test statistics work the other way around: a z or t statistic divides the observed difference by its standard error to ask how many SEs separate the data from the null hypothesis.

A useful mental shortcut follows: at 95% confidence, the interval is roughly the estimate ±2 standard errors. If a reported effect is smaller than about two of its own standard errors, it is not statistically distinguishable from zero.

Worked Example: Five Data Points to a Standard Error

Paste the values 4, 8, 6, 5, 12 into raw-data mode. The calculator performs each step exactly as you would by hand:

  1. Count and mean: n = 5 and the mean is (4 + 8 + 6 + 5 + 12) / 5 = 35 / 5 = 7.
  2. Squared deviations: (−3)² + 1² + (−1)² + (−2)² + 5² = 9 + 1 + 1 + 4 + 25 = 40.
  3. Sample variance: 40 / (5 − 1) = 10.
  4. Sample standard deviation: √10 ≈ 3.1623.
  5. Standard error: 3.1623 / √5 ≈ 1.4142.

So while individual values scatter about 3.2 units around the mean, the mean itself is uncertain by only about 1.4 units. (A tidy coincidence in this dataset: √10 / √5 = √2 ≈ 1.4142 exactly.)

Proportion mode is even quicker. If 40% of 100 surveyed customers renew their subscription: p(1 − p) = 0.4 × 0.6 = 0.24, divided by 100 gives 0.0024, and √0.0024 ≈ 0.049 — the renewal estimate is uncertain by about 4.9 percentage points per standard error.

Frequently Asked Questions

What is the difference between standard deviation and standard error?

Standard deviation measures how much individual observations spread around their mean; it describes the data. Standard error measures how much the sample mean itself would fluctuate across repeated samples; it describes the precision of an estimate. They are linked by SE = SD divided by the square root of n, so the standard error is always the smaller of the two for any sample larger than one observation.

Should the standard deviation use n or n - 1 when computing standard error?

Use the sample standard deviation with the n - 1 denominator (Bessel's correction). Dividing by n - 1 corrects the bias that arises because deviations are measured from the sample mean rather than the unknown population mean. This calculator's raw-data mode applies the correction automatically.

Why does the standard error get smaller as the sample grows?

Random errors of individual observations partially cancel when averaged, so the mean of many values is more stable than any single value. The cancellation follows the square root of n: four times the data gives half the standard error, one hundred times the data gives one tenth.

What counts as a small or acceptable standard error?

There is no universal threshold, because the standard error carries the units of the estimate. Judge it relative to the decision at hand: an SE of 2 points is excellent when means differ by 30 points and useless when they differ by 1. A common working criterion is whether the estimate is at least twice its standard error away from the value you want to rule out.

How do I turn a standard error into a confidence interval?

Multiply the standard error by the critical value for your confidence level and add and subtract the result from the estimate. For 95% confidence with a large sample, that is roughly the estimate plus or minus 1.96 standard errors. In the worked example, 7 plus or minus 1.96 times 1.4142 gives an interval of about 4.23 to 9.77.