Absolute and relative uncertainty across the range

MEASUREMENT UNCERTAINTY

Absolute and relative uncertainty across the range

Changing the units of expression does not justify extrapolating a performance claim.

Revised 30 September 2026. The original publication date and address are retained.

Synthetic model u(x) equals the square root of 1 squared plus 0.04x squared. Absolute standard uncertainty rises from 1.08 units at 10 to 4.12 at 100, while relative uncertainty falls from 10.8% to 4.1%.
Figure 1. An illustrative uncertainty function contains both a fixed absolute and a proportional contribution. Neither constant absolute uncertainty nor constant relative uncertainty describes the whole curve. Open full-size figure.

Two expressions of the same estimate

For a non-zero result x, relative standard uncertainty is u(x)/|x|. Multiply by 100 to express it as a percentage. If expanded uncertainty U is used instead, label the quantity consistently and state its coverage basis. A percentage is not inherently more comprehensive or more valid than an uncertainty expressed in the result’s units.

At a single result

At x = 100 units, a standard uncertainty of 4 units is a relative standard uncertainty of 4%. These express the same information. If k = 2 is justified, expanded uncertainty is 8 units or 8%; changing from u to U is a different operation from changing units to percent.

Across the measuring interval

A constant SD model predicts that relative uncertainty becomes larger as the result approaches zero. A constant CV model predicts that absolute uncertainty increases in proportion to the result. Real procedures may need a combination or a piecewise description. Select the model from suitable evidence, not from which expression looks tidier.

The synthetic curve uses u(x) = √[1² + (0.04x)²], with independent contributions in compatible units. At 10 units, u is about 1.08 units (10.8%); at 100 units, about 4.12 units (4.1%). These are illustrative numbers, not performance specifications for an assay.

At the low end

Relative uncertainty becomes unstable or undefined near zero. An absolute uncertainty or suitable coverage interval may communicate the estimate better. Detection capability and quantification claims need their own evidence; MU alone does not establish a limit of detection or quantification.

A useful reporting habit

  • Name the result, procedure and concentration range.
  • Label standard or expanded uncertainty explicitly.
  • State the model and its supported range.
  • Do not extrapolate a QC percentage to an unstudied decision level without justification.

Discover more from Pathology Uncertainty

Subscribe to get the latest posts sent to your email.

Leave a Reply

Discover more from Pathology Uncertainty

Subscribe now to keep reading and get access to the full archive.

Continue reading