Mathematics and physics
How to propagate measurement uncertainty in a density calculation
Write the result as a function of its inputs. Express each contribution in the result unit. For an approximately linear model with uncorrelated inputs, combine contributions in quadrature.
Gather the model, raw readings, certificates, units, replicates, and shared calibration information. Separate variation in readings from the sources of result uncertainty.
Use the stated coverage factor and interpretation to convert a certificate’s expanded uncertainty into a standard uncertainty before combining it with other contributions.
How to do it
Write the measurement model
For density, ρ = m/V. Align units and list each input estimate, source, standard uncertainty, and its basis.
Compute sensitivity coefficients
Evaluate partial derivatives at the input estimates. For density the mass coefficient is 1/V and the volume coefficient is −m/V². Multiply by the input standard uncertainty to get a result-unit contribution.
Handle correlations and nonlinearity
Shared calibration, temperature correction, or fit data can induce correlations. Use covariance terms where appropriate. For inadequate linear approximation, consider a suitable approach such as GUM Monte Carlo propagation.
Report combined and expanded uncertainty
Give the combined standard uncertainty u_c. For expanded uncertainty, give U = k·u_c and explain the choice of k. Report the model, units, leading contributions, and coverage interpretation.
Worked example
Synthetic example with m = 20.00 g and V = 10.00 cm³. The inputs are treated as uncorrelated and propagated to first order.
| Input | Standard uncertainty | Standard contribution to density |
|---|---|---|
| Mass | 0.10 g | 0.010 g/cm³ |
| Volume | 0.05 cm³ | 0.010 g/cm³ |
| Combined | Root sum of squares | About 0.014 g/cm³ |
The density is 2.000 g/cm³. With k = 2 the expanded uncertainty is about 0.028 g/cm³. Coverage depends on the distributions and degrees of freedom.
Common questions
Can I just add two percentages?
First establish whether they are standard uncertainties, bounds, or other quantities. Combine uncorrelated standard contributions in quadrature for this model. Correlation and conservative bounds need separate treatment.
Does k = 2 always give 95% coverage?
Coverage depends on distribution, degrees of freedom, and method. Report the actual factor and assumptions.
Verified sources
The official sources below were checked on 2026-11-13. Access conditions and update dates are rechecked before publication.
RigorMed workspace
Build my experimental uncertainty budget
Bring the model, input values, units, and component sources. Prepare a sensitivity and contribution table, including correlations that need review.
