Explained clearly 5 min read

How to Round Uncertainty Values: NIST GLP-9 vs GUM Recommendations

Round uncertainty to two significant digits by default, then match the measured value to the same decimal place. NIST GLP-9 and the GUM agree in principle, but GLP-9 gives more prescriptive laboratory rules.

Short Answer

Round uncertainty to two significant digits by default, then match the measured value to the same decimal place. NIST GLP-9 and the GUM agree in principle, but GLP-9 gives more prescriptive laboratory rules.

Rule Statement: The Two-Significant-Digit Default

The governing principle is simple: round the uncertainty first, then round the measured value to the same decimal place. The GUM states that it is usually sufficient to give the standard uncertainty u(y) to two significant digits (JCGM 100:2008, §7.2.6). NIST GLP-9, Rounding, adopts the same two-significant-digit default for laboratory reporting, while adding practical guidance for cases where the first significant digit is 1 or 2.

GUM 7.2.6: “The numerical values of the estimate y and its standard uncertainty u(y) or a coverage interval should not be given with an excessive number of digits. It is usually sufficient to give u(y) to two significant digits.”

NIST GLP-9 is more prescriptive for NIST Office of Weights and Measures laboratories: report the uncertainty to two significant digits, and report the result to the same decimal place. When the leading digit of the uncertainty is 1 or 2, retaining three significant digits may be allowed to avoid excessive rounding loss. This is consistent with NIST TN 1297 and common calibration practice. This site is a precision-and-rounding reference, not just a calculator; the rules below apply whether you compute by hand or with our significant figures calculator.

Convention Comparison Table: GUM vs NIST GLP-9

Aspect GUM (JCGM 100:2008) NIST GLP-9
Primary recommendation Usually sufficient to give u(y) to two significant digits. Round uncertainty to two significant digits for reporting.
Leading digit 1 or 2 No special rule stated in §7.2.6. Three significant digits may be retained when the first significant digit is 1 or 2.
Measured value Consistent number of digits; same decimal place as uncertainty. Round result to the same decimal place as the uncertainty.
Expanded uncertainty Apply same reporting logic to U = k uc. Round U after multiplying by coverage factor k.
Tie-breaking Not prescribed by GUM. Follow documented laboratory procedure; NIST rounding guidance applies.
Scope General metrology uncertainty expression. NIST OWM laboratory practice and legal metrology.

Worked Examples

Example 1: Two significant digits, leading digit 2

Suppose u = 0.0234 and y = 1.2344. The uncertainty has significant digits 2, 3, 4. Two significant digits gives 0.023. The uncertainty now has three decimal places, so round y to three decimal places: 1.234. Report: 1.234 ± 0.023.

Example 2: Small uncertainty with leading zeros

Let u = 0.001234 and y = 0.56789. Leading zeros are not significant. The first two significant digits are 1 and 2, so two significant digits gives 0.0012. This has four decimal places, so y becomes 0.5679. Report: 0.5679 ± 0.0012.

Example 3: NIST GLP-9 leading-digit exception

Let u = 0.0128 and y = 0.45678. Two significant digits would give 0.013. Because the first significant digit is 1, NIST GLP-9 permits three significant digits: 0.0128. Then y is rounded to four decimal places: 0.4568. Report: 0.4568 ± 0.0128. Under a strict two-digit rule, the same data would be reported as 0.457 ± 0.013.

Example 4: Expanded uncertainty

If uc = 0.000345 and k = 2, then U = 0.000690. Round U to two significant digits: 0.00069. The result y = 1.23456 is rounded to five decimal places: 1.23456. Report: 1.23456 ± 0.00069 (or with k = 2).

Counter-Examples

  • Too few digits: Rounding u = 0.0149 to one significant digit gives 0.01, discarding nearly half the uncertainty information. Two digits gives 0.015.
  • Too many digits: Reporting 12.3456 ± 0.02 is inconsistent. The uncertainty has two decimal places, so the value should be 12.35 ± 0.02.
  • Rounding the value first: If you round y before u, you may choose the wrong decimal place. Always round u first.
  • Double rounding: Rounding 0.0249 to 0.025 and then to 0.02 is wrong. Round once to two significant digits: 0.025.
  • Ignoring leading zeros: 0.00234 has three significant digits (2, 3, 4). Two significant digits is 0.0023, not 0.002.

Standards Citation

  • JCGM 100:2008 (GUM), §7.2.6: Recommends two significant digits for uncertainty and consistency between value and uncertainty.
  • NIST TN 1297, §7.2: NIST guidance on expressing uncertainty; supports two significant digits and the leading-digit 1/2 exception.
  • NIST GLP-9, Rounding: NIST OWM laboratory practice for rounding uncertainty and results.
  • ISO/IEC 17025:2017, §7.8.2.1: Requires reporting uncertainty in a way that is consistent with the result and documented procedure.
  • ASTM E29: Standard practice for using significant digits in test data to determine conformance.
  • ISO 80000-1:2009, Clause 7.3: General rules for rounding numerical values and expressing quantities.

Common Mistakes

  • Reporting uncertainty with more than two significant digits when the leading digit is 3 or greater.
  • Using one significant digit for all uncertainties “to be safe.” This overstates precision loss.
  • Mixing significant digits with decimal places. The uncertainty determines the decimal place of the result.
  • Forgetting that the coverage factor k must be applied before rounding the expanded uncertainty.
  • Using software default rounding without checking binary floating-point behavior. See our JavaScript floating-point note.
  • Assuming GUM and NIST GLP-9 are identical. GUM is general; GLP-9 is an operational NIST laboratory rule set.

Quick Reference Table

Raw uncertainty Two significant digits Three digits if leading 1 or 2 Result decimal places
0.0234 0.023 — 3
0.001234 0.0012 0.00123 4 or 5
0.0128 0.013 0.0128 3 or 4
0.000345 0.00035 — 5
0.9876 0.99 — 2
1.2345 1.2 — 1

Practice Problems

  1. Round u = 0.04567 and y = 12.3456. Answer: u = 0.046; y = 12.346; report 12.346 ± 0.046.
  2. Round u = 0.0001234 and y = 0.98765. Answer: two digits: u = 0.00012; y = 0.98765. With GLP-9 leading-digit exception: u = 0.000123; y = 0.987650.
  3. Round u = 0.9876 and y = 123.456. Answer: u = 0.99; y = 123.46; report 123.46 ± 0.99.

Sources & Further Reading

  • JCGM 100:2008, Evaluation of measurement data — Guide to the expression of uncertainty in measurement (GUM), §7.2.6.
  • NIST Technical Note 1297, Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results, §7.2.
  • NIST GLP-9, Rounding, NIST Office of Weights and Measures.
  • ISO/IEC 17025:2017, General requirements for the competence of testing and calibration laboratories, §7.8.2.1.
  • ASTM E29, Standard Practice for Using Significant Digits in Test Data to Determine Conformance with Specifications.
  • ISO 80000-1:2009, Quantities and units — Part 1: General, Clause 7.3.

FAQ

Should I always round uncertainty to two significant digits?

The GUM says two significant digits are usually sufficient. NIST GLP-9 generally requires two, but permits three when the first significant digit is 1 or 2. For high-precision intermediate work, retain more digits and round only at reporting.

Do I round the measured value before or after the uncertainty?

Round the uncertainty first. Then round the measured value to the same decimal place as the rounded uncertainty. This prevents inconsistent reporting.

Does NIST GLP-9 conflict with the GUM?

No. GLP-9 is consistent with the GUM’s principles but adds prescriptive laboratory rules for NIST OWM applications. When a contract or accreditation body requires a specific standard, follow that requirement and document it.

When can I keep three significant digits for uncertainty?

NIST GLP-9 and NIST TN 1297 allow three significant digits when the first significant digit of the uncertainty is 1 or 2. This reduces rounding loss while still limiting excessive digits.

Verified sources

References

  1. JCGM 100:2008, Evaluation of measurement data — Guide to the expression of uncertainty in measurement (GUM), §7.2.6.
  2. NIST Technical Note 1297, Guidelines for Evaluating and Expressing the Uncertainty of NIST Measurement Results, §7.2.
  3. NIST GLP-9, Rounding, NIST Office of Weights and Measures.
  4. ISO/IEC 17025:2017, General requirements for the competence of testing and calibration laboratories, §7.8.2.1.
  5. ASTM E29, Standard Practice for Using Significant Digits in Test Data to Determine Conformance with Specifications.

Leave a Reply

Your email address will not be published. Required fields are marked *