Uncertainty Propagation Calculator
Combine independent measurement uncertainties in a weighted sum, product or ratio. See which input contributes most and explore how improving it changes the result.
See what drives measurement uncertainty.
Build a measurement budget. Find the largest contributors and explore which input to improve.
Your measurement model
3 of 12 inputs. Each uncertainty is assumed uncorrelated with the others.
Multiplier for this input in the sum. Use −1 for a subtraction.
Input name and units
Labels do not convert values. Enter estimates and uncertainties in matching units.
Output and coverage settings
Choose a label consistent with your model and sensitivity coefficients.
Expanded uncertainty U = k × combined standard uncertainty. A factor alone does not establish a confidence percentage.
A fixed value added after the weighted sum. Put an uncertain correction in its own input row.
Estimate ± expanded uncertainty
First-order estimate, zero covariance. Coverage factor k = 2; no confidence percentage is inferred.
- Combined standard uncertainty uc
- ≈0.127017 mm
- Relative standard uncertainty
- ≈0.12689 %
- Expanded interval, y − U to y + U
- ≈99.846 mmto≈100.354 mm
Shares of first-order output variance, ordered from largest to smallest. Choose an input to inspect a reduction in its entered uncertainty. Equal contributions keep their original order.
What would a better measurement change?
Reduce the uncertainty of Reading while keeping all estimates and other contributions fixed. This study leaves your entered values unchanged.
Combined standard uncertainty (mm)
Reduction in the selected input uncertainty (%)
- Baseline uc
- ≈0.127017 mm
- Inspected uc
- ≈0.127017 mm
- Inspected expanded U
- ≈0.254034 mm
- Reduction in combined uncertainty
- 0 %
Removing this input's first-order contribution entirely would reduce combined uncertainty by ≈67.2223 %. Other contributions remain.
- The study changes only the selected input's entered uncertainty; estimates, sensitivities and all other inputs stay fixed.
- Reducing an input uncertainty by 100% removes only its first-order contribution. This is a sensitivity study, not a forecast of attainable improvement.
- Very small contributions and improvements can be difficult to distinguish visually; use the exact values and study table.
All 51 improvement samples
| Input uncertainty reduction | Retained input uncertainty | Combined variance | Combined standard uncertainty | Expanded uncertainty | Combined uncertainty reduction |
|---|---|---|---|---|---|
| 0.12 mm | ≈0.0161333 (mm)^2 | ≈0.127017 mm | ≈0.254034 mm | 0 % | |
| 0.1176 mm | ≈0.0155631 (mm)^2 | ≈0.124752 mm | ≈0.249504 mm | ≈1.78317 % | |
| 0.1152 mm | ≈0.0150044 (mm)^2 | ≈0.122492 mm | ≈0.244985 mm | ≈3.56229 % | |
| 0.1128 mm | ≈0.0144572 (mm)^2 | ≈0.120238 mm | ≈0.240476 mm | ≈5.33714 % | |
| 0.1104 mm | ≈0.0139215 (mm)^2 | ≈0.117989 mm | ≈0.235979 mm | ≈7.10746 % | |
| 0.108 mm | ≈0.0133973 (mm)^2 | ≈0.115747 mm | ≈0.231494 mm | ≈8.87299 % | |
| 0.1056 mm | ≈0.0128847 (mm)^2 | ≈0.113511 mm | ≈0.227022 mm | ≈10.6335 % | |
| 0.1032 mm | ≈0.0123836 (mm)^2 | ≈0.111282 mm | ≈0.222563 mm | ≈12.3885 % | |
| 0.1008 mm | ≈0.011894 (mm)^2 | ≈0.109059 mm | ≈0.218119 mm | ≈14.1379 % | |
| 0.0984 mm | ≈0.0114159 (mm)^2 | ≈0.106845 mm | ≈0.21369 mm | ≈15.8812 % | |
| 0.096 mm | ≈0.0109493 (mm)^2 | ≈0.104639 mm | ≈0.209278 mm | ≈17.6181 % | |
| 0.0936 mm | ≈0.0104943 (mm)^2 | ≈0.102442 mm | ≈0.204883 mm | ≈19.3481 % | |
| 0.0912 mm | ≈0.0100508 (mm)^2 | ≈0.100254 mm | ≈0.200507 mm | ≈21.0708 % | |
| 0.0888 mm | ≈0.00961877 (mm)^2 | ≈0.0980753 mm | ≈0.196151 mm | ≈22.7857 % | |
| 0.0864 mm | ≈0.00919829 (mm)^2 | ≈0.0959077 mm | ≈0.191815 mm | ≈24.4922 % | |
| 0.084 mm | ≈0.00878933 (mm)^2 | ≈0.0937514 mm | ≈0.187503 mm | ≈26.1899 % | |
| 0.0816 mm | ≈0.00839189 (mm)^2 | ≈0.0916073 mm | ≈0.183215 mm | ≈27.878 % | |
| 0.0792 mm | ≈0.00800597 (mm)^2 | ≈0.0894761 mm | ≈0.178952 mm | ≈29.5558 % | |
| 0.0768 mm | ≈0.00763157 (mm)^2 | ≈0.0873589 mm | ≈0.174718 mm | ≈31.2227 % | |
| 0.0744 mm | ≈0.00726869 (mm)^2 | ≈0.0852566 mm | ≈0.170513 mm | ≈32.8778 % | |
| 0.072 mm | ≈0.00691733 (mm)^2 | ≈0.0831705 mm | ≈0.166341 mm | ≈34.5202 % | |
| 0.0696 mm | ≈0.00657749 (mm)^2 | ≈0.0811017 mm | ≈0.162203 mm | ≈36.1489 % | |
| 0.0672 mm | ≈0.00624917 (mm)^2 | ≈0.0790517 mm | ≈0.158103 mm | ≈37.7629 % | |
| 0.0648 mm | ≈0.00593237 (mm)^2 | ≈0.0770219 mm | ≈0.154044 mm | ≈39.361 % | |
| 0.0624 mm | ≈0.00562709 (mm)^2 | ≈0.075014 mm | ≈0.150028 mm | ≈40.9418 % | |
| 0.06 mm | ≈0.00533333 (mm)^2 | ≈0.0730297 mm | ≈0.146059 mm | ≈42.504 % | |
| 0.0576 mm | ≈0.00505109 (mm)^2 | ≈0.071071 mm | ≈0.142142 mm | ≈44.0461 % | |
| 0.0552 mm | ≈0.00478037 (mm)^2 | ≈0.0691402 mm | ≈0.13828 mm | ≈45.5662 % | |
| 0.0528 mm | ≈0.00452117 (mm)^2 | ≈0.0672397 mm | ≈0.134479 mm | ≈47.0625 % | |
| 0.0504 mm | ≈0.00427349 (mm)^2 | ≈0.065372 mm | ≈0.130744 mm | ≈48.5329 % | |
| 0.048 mm | ≈0.00403733 (mm)^2 | ≈0.06354 mm | ≈0.12708 mm | ≈49.9752 % | |
| 0.0456 mm | ≈0.00381269 (mm)^2 | ≈0.061747 mm | ≈0.123494 mm | ≈51.3868 % | |
| 0.0432 mm | ≈0.00359957 (mm)^2 | ≈0.0599964 mm | ≈0.119993 mm | ≈52.7651 % | |
| 0.0408 mm | ≈0.00339797 (mm)^2 | ≈0.0582921 mm | ≈0.116584 mm | ≈54.1068 % | |
| 0.0384 mm | ≈0.00320789 (mm)^2 | ≈0.0566383 mm | ≈0.113277 mm | ≈55.4089 % | |
| 0.036 mm | ≈0.00302933 (mm)^2 | ≈0.0550394 mm | ≈0.110079 mm | ≈56.6677 % | |
| 0.0336 mm | ≈0.00286229 (mm)^2 | ≈0.0535004 mm | ≈0.107001 mm | ≈57.8794 % | |
| 0.0312 mm | ≈0.00270677 (mm)^2 | ≈0.0520267 mm | ≈0.104053 mm | ≈59.0396 % | |
| 0.0288 mm | ≈0.00256277 (mm)^2 | ≈0.0506238 mm | ≈0.101248 mm | ≈60.1441 % | |
| 0.0264 mm | ≈0.00243029 (mm)^2 | ≈0.049298 mm | ≈0.098596 mm | ≈61.1879 % | |
| 0.024 mm | ≈0.00230933 (mm)^2 | ≈0.0480555 mm | ≈0.096111 mm | ≈62.1661 % | |
| 0.0216 mm | ≈0.00219989 (mm)^2 | ≈0.046903 mm | ≈0.093806 mm | ≈63.0734 % | |
| 0.0192 mm | ≈0.00210197 (mm)^2 | ≈0.0458473 mm | ≈0.0916946 mm | ≈63.9046 % | |
| 0.0168 mm | ≈0.00201557 (mm)^2 | ≈0.0448951 mm | ≈0.0897903 mm | ≈64.6542 % | |
| 0.0144 mm | ≈0.00194069 (mm)^2 | ≈0.0440533 mm | ≈0.0881066 mm | ≈65.317 % | |
| 0.012 mm | ≈0.00187733 (mm)^2 | ≈0.0433282 mm | ≈0.0866564 mm | ≈65.8879 % | |
| 0.0096 mm | ≈0.00182549 (mm)^2 | ≈0.0427258 mm | ≈0.0854516 mm | ≈66.3622 % | |
| 0.0072 mm | ≈0.00178517 (mm)^2 | ≈0.0422513 mm | ≈0.0845026 mm | ≈66.7357 % | |
| 0.0048 mm | ≈0.00175637 (mm)^2 | ≈0.0419091 mm | ≈0.0838182 mm | ≈67.0051 % | |
| 0.0024 mm | ≈0.00173909 (mm)^2 | ≈0.0417024 mm | ≈0.0834049 mm | ≈67.1678 % | |
| 0 mm | ≈0.00173333 (mm)^2 | ≈0.0416333 mm | ≈0.0832666 mm | ≈67.2223 % |
Retained input uncertainty keeps its entered basis: standard, half-width or expanded. Select a row to inspect it on the curve.
Complete uncertainty budget
Contributions add as variances. Ranks compare their exact values; a zero total has no defined percentage shares.
| Input | Estimate | Standard u | Sensitivity | Contribution variance | Share | Rank |
|---|---|---|---|---|---|---|
| Standard uncertainty · u | 100 mm | 0.12 mm | 1 (mm)/(mm) | 0.0144 (mm)^2 | ≈89.2562 % | 1 · largest |
| Expanded uncertainty · U | 0.2 mm | 0.03 mm | 1 (mm)/(mm) | 0.0009 (mm)^2 | ≈5.57851 % | 2 |
| Rectangular · half-width a | -0.1 mm | ≈0.0288675 mm | 1 (mm)/(mm) | ≈0.000833333 (mm)^2 | ≈5.16529 % | 3 |
| Combined variance | ≈0.0161333 (mm)^2 | |||||
Unit labels describe your entries. Sensitivities and the output unit must be consistent with the chosen measurement model.
Save or import inputs
Complete report
Save the full input budget, sensitivities, uncertainty contributions and all 51 improvement samples, with units and assumptions.
Complete report CSV
Method and limits
Each contribution is its input variance multiplied by the square of its sensitivity coefficient. With zero covariance, the contributions add to the combined variance. Its square root is the combined standard uncertainty.
The calculation uses the first-order method in NIST TN 1297 (1994). Product and ratio results are local approximations; large uncertainty and a denominator near zero can make them misleading. Correlation, higher-order effects and coverage-probability estimation are outside this model.
A rectangular or triangular entry is a symmetric half-width around the estimate. Expanded input uncertainty requires its stated multiplier. Constant offsets are treated as exact; put uncertain corrections into their own input rows.
Enter at most 12 significant digits and six decimal places, including scientific notation. Estimates, coefficients and uncertainty magnitudes are bounded to 10¹²; coverage factors are greater than zero and at most 100. Very small or large derived values can produce an explicit numerical-limit result. Inactive drafts are kept until their model or basis becomes active again.
Displayed values are rounded; ≈ marks approximation. The estimate ± uncertainty form can remain useful when rounded interval endpoints appear identical. Zero variance has no percentage budget, and a zero estimate has no relative uncertainty.
- First-order propagation uses the stated model and local sensitivity coefficients.
- Input quantities are treated as uncorrelated: all covariance terms are zero.
- Unit labels describe the entered basis; no dimensional conversion or consistency check is performed.
- Expanded uncertainty is coverage factor times combined standard uncertainty; no confidence probability is inferred.
How to use this calculator
- Choose Calculation model: a weighted sum, product or ratio. Add up to twelve inputs; a ratio requires exactly two, with the numerator first and a nonzero denominator second. Changing the model retains every row and inactive draft.
- Choose Selected input and enter its Estimate, Uncertainty basis and uncertainty amount. Standard uncertainty u is a standard-deviation scale; Half-width a is the symmetric bound for rectangular or triangular input; Expanded uncertainty U needs its stated Input coverage factor.
- For a weighted sum, enter each Sensitivity coefficient and the Constant offset. Review Input name, Input unit label, Output unit label and Output coverage factor k. Unit labels describe your chosen basis; they do not convert quantities or verify dimensional consistency.
- Read the nominal result, combined standard uncertainty and expanded interval. Compare the ranked contribution chart with Complete uncertainty budget. Zero variance makes shares unavailable; a zero nominal result makes relative uncertainty unavailable. Review any nonlinear or denominator warning.
- Select an input for the improvement study and inspect its reduction from 0% to 100%. The plot changes only that input's uncertainty contribution in the hypothetical study. Removing its first-order contribution does not establish that the physical measurement has no other uncertainty.
- Use Save or import inputs, Input file format and Input text to preview to stage a complete CSV or TSV plan, then apply it explicitly. Reusable input preserves unfinished drafts. Complete report is available only for a valid budget and study and includes raw inputs, assumptions, the ledger and all study samples.
The Formula
The mathematical logic powering this calculation is:
For independent input estimates xᵢ with standard uncertainties uᵢ, first-order variance is V = Σsᵢ²uᵢ², where sᵢ = ∂f/∂xᵢ at the entered estimates. Covariances are assumed zero. Combined standard uncertainty uc = √V; expanded uncertainty U = kuc; the displayed expanded interval is y − U to y + U. A coverage factor alone does not establish a confidence percentage.
Weighted sum: y = offset + Σcᵢxᵢ and sᵢ = cᵢ. Product: y = Πxᵢ and sᵢ is the product of all other estimates, including zero cases. Ratio: y = x₁/x₂, s₁ = 1/x₂ and s₂ = −x₁/x₂². A zero denominator is invalid. Product and ratio propagation are nonlinear approximations.
Standard input uses u² directly; a symmetric rectangular half-width a has variance a²/3; a symmetric triangular half-width a has variance a²/6; expanded input Uᵢ with stated factor kᵢ has variance (Uᵢ/kᵢ)². These distributions are entered assumptions, not inferred from the numbers.
Contribution variance vᵢ = sᵢ²uᵢ² and contribution share = vᵢ/V when V > 0. Relative uncertainty uses |y| and is unavailable when y = 0. For an illustrative reduction r in one selected input's entered uncertainty, V(r) = Σother vᵢ + vselected(1 − r/100)². The 51 study points leave the original estimates and input plan unchanged.Worked Example
If you input:
- Calculation model: Weighted sum
- Constant offset: 0 mm
- Reading: Estimate 100 mm; standard uncertainty u = 0.12 mm; sensitivity coefficient 1
- Calibration correction: Estimate 0.2 mm; expanded uncertainty U = 0.06 mm; input coverage factor 2; sensitivity coefficient 1
- Drift allowance: Estimate −0.1 mm; rectangular half-width a = 0.05 mm; sensitivity coefficient 1
- Output unit label: mm
- Output coverage factor k: 2
Result:
The nominal result is exactly 100.1 mm. Independent variance contributions sum to 121/7500 mm², so combined standard uncertainty is approximately 0.127017059222 mm and expanded uncertainty at k = 2 is approximately 0.254034118443 mm. The expanded interval is approximately [99.8459658816, 100.3540341184] mm. Reading contributes 108/121 of the variance, approximately 89.2562%. This illustrative k = 2 interval is not automatically a 95% confidence interval.
Frequently Asked Questions
What does independent uncertainty propagation assume?
This workspace sets all covariance terms to zero and evaluates sensitivities at the entered estimates. Inputs that share a calibration source, environment or calculation may be correlated even when they appear in separate rows. The calculator cannot establish independence from labels or numbers. It does not estimate correlations, fit a measurement model or include omitted sources of uncertainty automatically.
Which uncertainty amount should I enter?
Standard means an uncertainty expressed on a standard-deviation scale. Rectangular and triangular mean a symmetric half-width around the estimate, converted using a/√3 and a/√6 respectively. Expanded means an already expanded uncertainty with its stated input coverage factor, divided by that factor. A tolerance or certificate interval should not be relabeled as standard uncertainty without a justified distribution or coverage factor.
Does output coverage factor 2 mean 95% confidence?
Not by itself. The output factor multiplies combined standard uncertainty to produce an expanded uncertainty. Relating that interval to a coverage probability requires additional distribution and uncertainty-estimation information. This tool does not calculate effective degrees of freedom or claim automatic 95% coverage, and its expanded interval is different from a statistical confidence interval for a population mean.
Why are some zero results accompanied by warnings?
A zero nominal result makes relative uncertainty undefined even when absolute uncertainty is useful. With zero total variance, contribution shares and percentage improvement are unavailable. A product with two or more zero estimates can have zero first-order sensitivities while still varying at higher order. Its explicit warning prevents that algebraic result from being interpreted as proof of zero actual uncertainty.
What does the ratio denominator warning mean?
The warning appears when the denominator's estimate plus or minus the output coverage factor times its standard uncertainty contains zero, including equality. A symmetric first-order interval may then be misleading. A denominator estimate of exactly zero is invalid; a nonzero denominator with this warning still produces the stated approximation. The workspace does not replace it with a bounded exact ratio distribution.
Can the improvement chart recommend a better instrument?
It shows how reducing one entered uncertainty changes the first-order budget while keeping nominal estimates and other contributions fixed. The selected row, reduction and inspection are view settings and do not overwrite the authored input. The chart does not estimate cost, feasibility, instrument performance or unmodeled uncertainty. Its 100% endpoint removes only the selected contribution, not every source in a physical measurement.
What happens to inactive drafts, units and numerical limits?
Inactive coefficients, offset and input factors remain unchanged and are ignored numerically until active. Unknown model or distribution tokens need explicit repair. Units are descriptive labels with no automatic conversion. Active numbers allow 12 significant digits and six effective decimal places; supported published nonzero quantities lie between 10⁻³⁰⁰ and 10³⁰⁰. Invalid input or a numerical limit prevents a complete report rather than substituting zero or a stale result.
Similar Calculators
Standard Deviation Calculator
Calculate sample or population standard deviation, variance, mean and extrema from complete decimal data. Preserve small differences and identify invalid entries.
Confidence Interval Calculator
Calculate a two-sided confidence interval for a mean using sample SD with Student t or known population SD with z. See the margin and standard error.
Ratio Calculator
Simplify ratios, find equivalent ratios, and solve for a missing value in a proportion.
Pitching Speed Calculator & Video Timing
Estimate average baseball pitch speed from video frame marks or elapsed time. Compare timing bounds and find the capture rate for your desired precision.
