Questions: Uncertainty in Analytical Measurement

Question 1

A water testing laboratory measures a lead concentration of 9 ppb. The regulatory limit is 10 ppb, and the measurement's expanded uncertainty is ±3 ppb at 95% confidence. Can the laboratory certify that this sample complies with the regulation?

• A. Yes — the measured value (9 ppb) is below the regulatory limit (10 ppb)
• B. No — the expanded uncertainty means the true concentration could plausibly range from 6 to 12 ppb, including values above the limit
• C. Yes — a 95% confidence level is the standard for regulatory compliance and is sufficient here
• D. No — compliance decisions require uncertainty below ±1 ppb regardless of the regulatory limit

Answer: (B)

The whole point of reporting uncertainty is to communicate the range within which the true value plausibly lies. A result of 9 ± 3 ppb means the true value could be as high as 12 ppb — well above the 10 ppb limit. Reporting only the central estimate and comparing it to the limit, while ignoring uncertainty, gives a false sense of precision. Regulatory frameworks specifically require uncertainty estimation so that borderline results are not falsely certified as compliant. The laboratory cannot certify compliance without reducing its measurement uncertainty.

Question 2

A laboratory's error budget analysis shows that 78% of its total measurement uncertainty comes from the field sampling step. What is the most effective way to reduce the total uncertainty?

• A. Purchase a higher-precision analytical instrument, since instrument noise is often the dominant error source
• B. Improve the sampling protocol — since sampling dominates the budget, reducing its contribution will most significantly reduce total uncertainty
• C. Increase the number of replicate analyses per sample to reduce the standard error of the mean
• D. Recalibrate the instrument more frequently to reduce calibration drift

Answer: (B)

The error budget exists precisely to direct improvement efforts. When total uncertainty is dominated by one source (here, sampling), reducing other sources — even to zero — barely changes the total, because uncertainty combines as root-sum-of-squares. If sampling contributes 78% of variance, the remaining sources together contribute only 22%. Buying a better instrument that halves instrumental uncertainty would barely move the total. The leverage is entirely on the dominant source. This is a counterintuitive and practically important lesson: analytical precision in the lab is often irrelevant when field sampling is sloppy.

Question 3

Reporting a measurement result without an associated uncertainty is incomplete, because the reported number alone conveys no information about how reliable or precise it is.

• T. True
• F. False

Answer: True

A single number like '15 ppb' could represent a result reproducible to ±0.1 ppb or one reproducible only to ±5 ppb — you cannot tell from the number alone. Uncertainty quantifies the range within which the true value plausibly lies, given all sources of error in the measurement process. Without this range, the result cannot be used for regulatory compliance, quality control, or scientific comparison. The GUM framework exists precisely to standardize how this uncertainty is estimated and reported, making results interpretable by anyone who receives them.

Question 4

Reducing the uncertainty of each individual source in an error budget by 50% will reduce the total combined uncertainty by approximately 50%.

• T. True
• F. False

Answer: False

Uncertainty sources combine as root-sum-of-squares (for independent contributions), not as a simple sum. If one source dominates — say, contributing 80% of the total variance — then reducing all other sources to zero reduces total uncertainty by only about 10%, because the dominant source still determines the bulk of the combined value. Conversely, halving the dominant source reduces total uncertainty much more than halving a minor source. The nonlinear nature of RSS combination means that targeted reduction of the largest contributor is far more effective than across-the-board reductions.

Question 5

Why is a measurement result without an associated uncertainty considered incomplete, and how does an error budget help a laboratory improve its results?

Model answer: Without uncertainty, a measurement result cannot be used to make any decision that depends on knowing whether the true value falls within some range — regulatory compliance, product release, scientific comparison. The number alone is a point estimate; the uncertainty gives it context. An error budget improves results by decomposing total uncertainty into its component sources (sampling, preparation, calibration, instrument, environment), revealing which steps contribute the most error. Because improvement resources are limited, knowing that one step dominates the budget directs effort where it will have maximum impact — often revealing that buying better instruments is irrelevant when field sampling is the true bottleneck.

The GUM framework formalizes this: estimate standard uncertainty for each component, combine using root-sum-of-squares, then multiply by a coverage factor (typically k = 2) for the expanded uncertainty at ~95% confidence. The final result is reported as x ± U with coverage factor stated. Regulated industries require this not as bureaucratic formality but because it is the only way to rigorously assess whether a measurement result is fit for its intended purpose.