Accounting & Integrity
Error propagation
The mathematical combination of individual uncertainties into the uncertainty of a calculated result.
When quantities are multiplied, relative uncertainties add in quadrature; when added, absolute uncertainties combine. Propagation turns a list of input ranges into a defensible interval around the output.
The method assumes independence. Where inputs are correlated — as when the same sampling design produces several of them — naive propagation understates the total, and a Monte Carlo approach is preferable.
Frequently asked questions
Why quadrature rather than simple addition?
Because independent random errors partially cancel, so variances add rather than standard deviations.
When is simulation needed?
When relationships are non-linear or inputs are correlated, which is common in carbon accounting.