Choose your reading route
eduKateSG · Why Mathematics? · Normalised Error Eₙ, Combined Uncertainty and Measurement Comparisons
Start with the question, audit the worked example, or jump to the assumptions and practice. Every chapter stays open for uninterrupted reading.
Understand the question
Define the quantity and follow the mechanism.Route 2 · Chapters 4–7
Reproduce the example
Audit the numbers one step at a time.Route 3 · Chapters 8–13
Check assumptions
See where design and calibration matter.Route 4 · Chapters 14–18
Explore applications
Connect the method to a concrete investigation.Route 5 · Chapters 19–23
Study and verify
Practise, inspect failure cases and read further.
Full chapter index · Practice and checking · Mathematics Learning Hub
Two laboratories can report different values and different uncertainties. Normalised error, written E_n, asks whether their difference is large relative to the combined stated uncertainty. It is a compact calculation used in measurement comparisons and proficiency contexts, especially where an assigned reference value and expanded uncertainties are available.
The importance of mathematics lies in the denominator: subtraction alone tells us the gap, while root-sum-of-squares places that gap on an uncertainty scale. The result supports a defined comparison; it does not prove universal competence or identify the cause of disagreement.
Full chapter index
Chapters 1–3 · Understand the question
Chapters 4–7 · Reproduce the example
Chapters 8–13 · Check assumptions
Chapters 14–18 · Explore applications
Chapters 19–23 · Study and verify
Let x be a participant result and X the assigned comparison value. Their signed difference is x – X. Let U_lab and U_ref be the corresponding expanded uncertainties stated at compatible coverage levels.
The normalised error is E_n = (x – X) divided by square root of U_lab squared plus U_ref squared. The sign gives direction; the absolute value gives the comparison with the combined uncertainty scale.
CHAPTER 2 OF 23 · Understand the question
2. Why uncertainties combine by root-sum-of-squares
If two uncertainty contributions are independent, the variance of a difference is the sum of the variances. Standard deviations therefore combine through a square root of summed squares. The same idea appears in vector length and the Pythagorean theorem.
Correlation changes the calculation. If the participant and reference values share standards or calibrations, covariance must be considered under the applicable comparison model. Independence is an assumption, not punctuation.
A participant reports x = 100.8 units with expanded uncertainty U_lab = 0.8 unit. The assigned value is X = 100.0 units with U_ref = 0.6 unit. Both uncertainties use the same stated coverage convention.
The numerator is 100.8 – 100.0 = 0.8 unit. The denominator is square root of 0.8 squared plus 0.6 squared = square root of 0.64 + 0.36 = square root of 1 = 1.0 unit.
| Quantity | Value | Role |
|---|---|---|
| Participant result x | 100.8 | Reported result |
| Assigned value X | 100.0 | Comparison reference |
| U_lab | 0.8 | Participant expanded uncertainty |
| U_ref | 0.6 | Reference expanded uncertainty |
E_n = 0.8/1.0 = 0.8. A BIPM-hosted metrology comparison report states the commonly used criterion |E_n| ≤ 1 for satisfactory comparison performance and |E_n| > 1 for unsatisfactory performance in that context.
The correct sentence is narrow: this result is within the combined expanded uncertainty criterion used for the comparison. It is not a claim that every future measurement will be correct.
| Step | Calculation | Result |
|---|---|---|
| Difference | 100.8 − 100.0 | 0.8 |
| Combined scale | √(0.8² + 0.6²) | 1.0 |
| En | 0.8 / 1.0 | 0.8 |
If x were 99.2 with the same uncertainty, E_n would be -0.8. The sign switches because the participant value is below the assigned value, but |E_n| remains 0.8.
Keep the signed value in reports because direction can reveal a consistent high or low pattern across points. Use the absolute value only for the threshold comparison.
If x = 101.3 while X and both uncertainties stay unchanged, the numerator is 1.3 and E_n = 1.3. This exceeds the illustrative criterion. The next action is an investigation, not an invented correction.
Check units, transcription, environmental corrections, calibration chains, uncertainty budgets and possible correlations. A large E_n can result from a shifted value, understated uncertainty or both.
Do not mix a standard uncertainty with an expanded uncertainty in the same formula. Expanded uncertainty is commonly written U = ku_c, where u_c is combined standard uncertainty and k is a coverage factor. The comparison protocol should specify the convention.
The BIPM-hosted example used expanded uncertainties with k = 2. If coverage factors differ, follow the relevant procedure rather than forcing unlike numbers into one denominator.
A larger reported uncertainty increases the denominator and can reduce |E_n|, but that is not permission to inflate uncertainty. Each budget must be technically justified, complete and traceable. An unrealistic budget weakens the comparison.
A very small uncertainty is not automatically admirable either. It must cover relevant contributions and be consistent with observed performance.
Suppose both values depend on the same calibration standard. The shared contribution may cancel partly in the difference. A covariance-aware denominator can differ from the independent root-sum-of-squares expression.
Comparison reports should state how correlation is handled. Students should learn a general rule: identify shared inputs before adding uncertainties as if they were independent.
CHAPTER 10 OF 23 · Check assumptions
10. Assigned value and metrological traceability
An assigned value can come from a reference laboratory, a consensus process or another documented route. The way it is assigned affects its uncertainty and interpretation. E_n requires more than a number labelled correct.
Traceability connects a result through documented calibrations to references, with uncertainty along the chain. It does not eliminate variation.
One E_n value is one comparison. Across several measurement points, plot signed E_n against range, date or nominal value. A constant sign may reveal an offset; growth with magnitude may indicate scale error; isolated points may indicate local problems.
Do not count many correlated points as independent victories. The comparison design determines the evidence unit.
If an unrounded E_n is 1.004, displaying 1.00 can hide that it exceeds one. Keep enough digits for the decision and state the rounding rule. Calculate from unrounded inputs.
Boundary cases deserve calm review because uncertainty estimates themselves have limited knowledge. The threshold is a decision rule, not a physical cliff.
Two groups measure a common safe object and report results with justified uncertainty intervals. First calculate the difference. Then combine the uncertainty scales and compute a normalised discrepancy. Discuss what independence would mean in the classroom setup.
This analogy teaches the mechanism, but a classroom interval is not automatically a formal calibration uncertainty. Label it as an educational model.
Use columns for participant result, assigned value, U_lab, U_ref, covariance treatment and E_n. Validate that units match and uncertainties are non-negative. Calculate the denominator explicitly before the ratio so reviewers can audit each step.
Add conditional formatting only after the numerical logic is correct. Preserve the signed value even if the display highlights |E_n|.
A program should validate units and coverage conventions, calculate signed differences and denominators, flag zero denominators, and keep full precision. Plot results with point identifiers and the +1 and -1 reference lines.
Version the input file and uncertainty assumptions. Reproducible code cannot rescue undocumented metrology.
Coverage probability counts observations inside a fixed practical tolerance. E_n instead scales one comparison difference by stated uncertainties. The total deviation index estimates an individual agreement boundary from a distribution.
These tools answer different questions. Select the question before selecting the statistic.
|E_n| below one does not prove the participant uncertainty is perfect. |E_n| above one does not identify the cause. The statistic does not replace stability checks, method validation, competence assessment or technical review.
It also cannot be compared across exercises without understanding assigned values, uncertainty models and dependence.
Students can connect algebra, square roots, units and probability through a realistic example. Ask why the denominator has the same unit as the numerator and why the final ratio is dimensionless.
Parents can encourage precise language: say what comparison was made, what uncertainty convention was used and what the result supports.
Normalised error appears in calibration, electrical and dimensional metrology, laboratory comparisons and quality systems. The transferable skill is uncertainty reasoning. It supports engineering and science careers without guaranteeing entry or results.
The Mathematics Learning Hub offers connected school topics, and How Mathematics Works explains how models carry assumptions.
Recalculate E_n when x = 100.5. Then halve both uncertainties and explain why the same physical difference appears larger on the uncertainty scale. Convert every value and uncertainty from metres to millimetres and verify that E_n is unchanged.
Finally, add a positive covariance term under a documented alternative formula and describe why the denominator changes.
### Is E_n a percentage error? No. It is a dimensionless signed difference divided by a combined uncertainty scale.
### Why is one used as a boundary? Under the stated convention, one means the magnitude of the difference equals the combined expanded uncertainty. The governing programme defines its decision rule.
### Can I ignore the sign? Use |E_n| for a symmetric threshold, but retain the sign for diagnosis across points.
### Does a bigger uncertainty improve measurement? No. It may reduce |E_n|, but uncertainty must be evidence-based and fit the measurement model.
The BIPM-hosted SIM.EM-S15 final report gives the formula, the use of expanded uncertainties and its stated criterion in a real resistance comparison. The NIST metRology project situates comparison analysis within uncertainty methods. This article's numbers are original teaching data.
Normalised error shows how mathematics makes a disagreement interpretable. The numerator preserves direction, the denominator respects uncertainty, and the ratio supports a declared decision rule. Its value comes from honest inputs and a well-defined comparison.
Extended practice laboratory
Practice case 1: Move the participant result upward
Setup: Use the article’s normalised-error example and change one declared input for case 1. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 2: Move the result downward
Setup: Use the article’s normalised-error example and change one declared input for case 2. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 3: Double both uncertainties
Setup: Use the article’s normalised-error example and change one declared input for case 3. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 4: Halve both uncertainties
Setup: Use the article’s normalised-error example and change one declared input for case 4. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 5: Convert units
Setup: Use the article’s normalised-error example and change one declared input for case 5. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 6: Reverse the subtraction
Setup: Use the article’s normalised-error example and change one declared input for case 6. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 7: Round near one
Setup: Use the article’s normalised-error example and change one declared input for case 7. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 8: Set equal uncertainties
Setup: Use the article’s normalised-error example and change one declared input for case 8. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 9: Use an assigned value uncertainty of zero
Setup: Use the article’s normalised-error example and change one declared input for case 9. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 10: Introduce shared correlation
Setup: Use the article’s normalised-error example and change one declared input for case 10. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 11: Mix coverage factors
Setup: Use the article’s normalised-error example and change one declared input for case 11. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 12: Check a zero denominator
Setup: Use the article’s normalised-error example and change one declared input for case 12. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 13: Plot several signed values
Setup: Use the article’s normalised-error example and change one declared input for case 13. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 14: Find a constant offset
Setup: Use the article’s normalised-error example and change one declared input for case 14. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 15: Find a scale trend
Setup: Use the article’s normalised-error example and change one declared input for case 15. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 16: Investigate one exceedance
Setup: Use the article’s normalised-error example and change one declared input for case 16. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 17: Audit the reference value
Setup: Use the article’s normalised-error example and change one declared input for case 17. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 18: Audit the uncertainty budget
Setup: Use the article’s normalised-error example and change one declared input for case 18. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 19: Separate uncertainty from error
Setup: Use the article’s normalised-error example and change one declared input for case 19. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 20: Preserve the sign
Setup: Use the article’s normalised-error example and change one declared input for case 20. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 21: Use the absolute decision rule
Setup: Use the article’s normalised-error example and change one declared input for case 21. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 22: Check dependency
Setup: Use the article’s normalised-error example and change one declared input for case 22. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 23: Document expanded uncertainty
Setup: Use the article’s normalised-error example and change one declared input for case 23. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.
Practice case 24: Write a comparison-limited conclusion
Setup: Use the article’s normalised-error example and change one declared input for case 24. Task: Recalculate from the raw values, preserve units and state which design condition changed.
Check: Check arithmetic against the invariant described in the chapter, then separate the numerical finding from the practical claim. Rebuild the relevant counts before interpreting the coefficient, preserve the stated design, and write one sentence separating the numerical result from the claim it can support. Then vary one defensible assumption, calculate again, and explain whether the conclusion changes because of new data, a new reference model, or a new decision purpose. Finish by naming the unit, tolerance, and evidence needed for transfer. Record one graph or table that would expose failure hidden by the summary.