standard-error · shipping · also SE, Standard error of the mean, SEM

Standard error

The standard error of an estimator is the standard deviation of its sampling distribution under a specified sampling model.

For the arithmetic mean of n independent, identically distributed observations with finite population variance sigma^2, SE = sigma/sqrt(n); the usual estimate substitutes sample SD s. Normality is not needed for this variance identity. Correlated observations require accounting for their dependence. [1]

SE describes sampling variation, not total error, bias or the spread of individual latencies. Mean plus or minus one SE is not automatically a 95% confidence interval; the interval method and assumptions determine coverage. [2]

Hesela's invented 1-or-3 ms population has known SD 1 ms. Four independent draws have mean SE 0.5 ms. Copying each draw four times creates 16 records but leaves that SE at 0.5 ms. Treating those copies as independent would incorrectly give 0.25 ms using the known population SD, not an estimated sample SD. This is an exact counterexample, not a measured benchmark or a general effective-sample-size formula. [3]

Relations

related
Standard deviation, Confidence interval

Citations

  1. JCGM 100:2008 (GUM), 4.2.1-4.2.3 and 4.2.7: repeated observations and correlation (accessed 2026-11-19)
  2. NIST e-Handbook 1.3.5.2: Confidence Limits for the Mean (accessed 2026-11-19)
  3. Hesela: spread, mean uncertainty and copied latency records; exact synthetic example (accessed 2026-11-19)

JSON: /data/terms/standard-error.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. Standard error. Hesela. 2026-11-19. https://hesela.dev/terms/standard-error/