Usikkerhed - Uncertainty

Usikkerhed - Uncertainty

This page explains how uncertainty of chemical measurements is expressed by Eurofins Product Testing A/S in analytical reports.

Accredited testing and measurement always shows some uncertainty. Expression of that uncertainty has to follow guidance given in ISO 17025 and in documents of the International Laboratory Accreditation Cooperation (ILAC), such as ILAC-G17 "Introducing the Concept of Uncertainty of Measurement in Testing in Association with the Application of the Standard ISO/IEC 17025". The Danish accreditation authority DANAK concluded the requirements in its regulation "AB 13".

Uncertainty of chemical measurements

A series of measurements of the same object will deliver a certain "cloud" of results around a mean value. This will follow a normal distribution in many cases. Then the uncertainty of achieving a test result within this normal distribution can be expressed as RSD (relative standard deviation, RSD), and as expanded uncertainty, U_m:

The expanded uncertainty U_m (%) equals 2 x RSD%.

The factor 2 is the so-called coverage factor. If this factor equals 2 then this will correspond to a coverage probability of 95%. This means that if the same sample is measured 100 times, 95 of the results would be within the limits given by the mean value +/- the expanded uncertainty.

When a given method undergoes accreditation the uncertainty of measurement is determined by collecting data during several days, to ensure that the uncertainty covers worst case. It is common to assign conservative values to the measurement uncertainty, this means values are often given in numbers that can be divided by 5 (e.g. 5%, 10%, 15%).

In the analytical report the listed uncertainty is given for measurements higher than or equal to 5 x Limit of Detection (LOD). If the results reported are less than 5 x LOD, the measurement uncertainty can be even higher. It is commonly recognized that the measurement uncertainty can be up to a factor 5 higher if close to LOD level.

This is illustrated by the figure below. In this example, if the measurement uncertainty is 1% for measurements higher than 5 x LOD, then the uncertainty can grow to 5% at LOD level.

When an application for accreditation of a new analytical method is set up, this always is based on customer needs and not on the potential best performance of the method. This means that it would often be possible to achieve a lower LOD, than the one given in the analytical report.

If the true LOD is lower than reported, then the reported uncertainty is valid for the whole measurement range between true LOD and upper limit of measurement.