How this calculator works
RSE scales the standard error to the magnitude of the estimate. For an independent sample mean, SE is sample SD divided by sqrt(n). Survey weights, clustering, serial dependence or a different statistic may need another SE calculation; enter a valid externally calculated SE in summary mode in those cases. This tool uses the absolute estimate so negative estimates have a non-negative relative uncertainty.
Formula
RSE (%) = 100 × SE/|estimate|. For an independent mean, SE = s/sqrt(n). Conversely, SE = |estimate| × RSEratio.
Worked example
An estimate of 200 with SE 10 has RSE 5%. A sample mean of 100 with sample SD 20 and n = 100 has SE 2 and RSE 2%.
Advertisement
Frequently asked questions
Is a universal RSE threshold reliable?
No. Publication rules and acceptable uncertainty depend on the statistic, sampling design and intended use. This tool does not classify reliability.
Can I use observation SD as SE?
Only after applying the correct estimator-specific relationship. For independent sample means it is s/sqrt(n), not s alone.
What happens near a zero estimate?
RSE can become very large and is undefined at zero. Absolute SE may be a clearer description in that setting.