This article explores measurement uncertainty and its importance in modern lab work.
Voltair once wisely remarked, “Doubt is not a pleasant condition, but certainty is absurd .”
While no one wants to be unsure of their work, in science, 100% confidence brings down reliability! Acknowledging doubt or uncertainty is essential to create a culture of intellectual transparency in scientific research. It motivates researchers to seek alternative explanations, consider conflicting data, and continuously improve their processes.
Measurement uncertainty is the process of mathematically quantifying this doubt. It acknowledges that every measurement in the laboratory has error and uncertainty associated with the process itself.
Everything—from the accuracy of the measuring instrument to the day’s weather—creates variations in results.
What Is Measurement Uncertainty?
Measurement uncertainty is the inherent variability and lack of exactness in any measurement process. We cannot measure anything with absolute certainty due to various factors such as equipment limitations, environmental conditions, and human error.
The best we can do is estimate the range within which the actual value of a measured quantity lies.
There are two numbers we need to quantify measurement uncertainty. One is range or interval, and the other is confidence level.
For example, we may express biomarkers in the blood as “5.2 ± 0.3 ng/mL at a 95% confidence level.” or report air pollution levels as “40 ± 4 micrograms per cubic meter (µg/m³) at a 92% confidence level”. We mathematically express the value range of the results and their probability of occurrence. In other words, we express the result as the probability that the true value falls within a given range around the observed value.
Why Measurement Uncertainty Matters in the Laboratory
When it comes to measurement result, it is important to provide some quantitative indication of the quality of result obtained. This enables users to assess the measurement’s reliability and makes it easier to compare it with other measurements and reference values specified in standards and specifications.
That is what measurement uncertainty offers – a quantifiable measure of the result’s quality.
The concept of uncertainty as a quantifiable attribute represents a relatively recent development in the history of measurement. Despite error and error analysis being integral to the field of measurement science or metrology for quite some time, there is now a broad acknowledgment that even after evaluating all known or suspected error components and applying necessary corrections, a degree of uncertainty persists regarding the accuracy of the reported results.
From a practical standpoint, measurement uncertainty becomes critical when we aim to accomplish the following:
Comparing or reproducing results
Measurement uncertainty is critical when it comes to comparing or replicating results, a scenario often encountered during the verification and validation of a new test method..
Uncertainty analysis is an important component of method validation, especially when we are comparing the result of the new method with a well-established benchmark. This analysis serves as a critical step in assessing the various factors that influence the measurement process.
It acknowledges the inaccuracies inherent in instruments and their calibration to reference standards, as well as the human element involved in conducting the test, the environmental conditions, and the test method itself.
By understanding and accounting for these uncertainties, you are more able to to make informed judgments regarding the comparability of the new test method with an existing one and its ability to consistently reproduce results.
Providing a pass and fail result
In scenarios where pass or fail decisions carry substantial consequences, such as quality control and regulatory compliance, measurement uncertainty becomes an critical tool.
Rather than relying solely on a single data point, measurement uncertainty offers a comprehensive view by quantifying the range within which the true value of a measured parameter is likely to fall. This range takes into account both random and systematic errors inherent in the measurement process.
By utilising measurement uncertainty, you can establish appropriate thresholds for pass or fail decisions. This approach ensures that your findings are not only justifiable but also reliable, effectively reducing the risk of erroneous outcomes, whether they be false positives or false negatives.
Establishing standard processes
Quantifying uncertainty allows us to identify the primary sources of measurement error or variability. This knowledge is invaluable when developing standard testing procedures, enabling us to optimise conditions, choose appropriate instruments, and establish effective quality control measures
For example, in a water quality testing laboratory, uncertainty quantification helps identify factors affecting measurement reliability. With this insight, the lab can fine-tune testing conditions by adjusting parameters like sample volume, filtration methods, or incubation times. These refinements minimize uncertainty, leading to more accurate measurements and ensuring dependable results.
For example, a laboratory conducting water quality testing optimises testing conditions by adjusting parameters such as sample volume, filtration methods, or incubation times to minimize uncertainty and improve measurement accuracy.
Assessing risks for critical decision making
Measurement uncertainty is vital when decisions are based on scientific data obtained through measurement—such as in environmental monitoring, manufacturing, pharmaceutical production, or medical diagnosis. By accounting for uncertainty, we can evaluate potential risks with greater accuracy.
Measurement Uncertainty and ISO accreditation
If a laboratory wishes to gain ISO accreditation for their calibration or testing services, ISO/IEC 17025 General Requirements of the Competence of Testing and Calibration Laboratories” requires them to implement a procedure to estimate the uncertainty of measurement for all calibrations. It also requires them to provide calibration reports to the customer that contain the measurement results and a statement regarding the measurement uncertainty. Measurement results must also be traceable, meaning that it is traceable to a national standard through an unbroken chain of calibrations or comparisons, each having a stated uncertainty.
When Measurement Uncertainty Doesn’t Matter
Measurement uncertainty always matters in the context of scientific research and laboratory work. However, there are certain scenarios where the impact of measurement uncertainty is minimal.
For example, in the early stages of a research project, the focus is on generating initial data and exploring new ideas. Initially, the focus may be on feasibility, functionality, or proof of concept instead of precise measurements. Similarly, there may be research scenarios where qualitative descriptions are sufficient or have higher priority than quantitative measurements. For example, the behavioral study of animals or phenomenological studies, where researchers explore the lived experiences of individuals in response to specific situations.
However, as research matures toward production or other practical action, measurement uncertainty becomes more critical for quality control and meeting specifications.
How to Quantify Measurement Uncertainty
Measurement results have uncertainty because there are uncertainty sources (effects that cause uncertainty). These are effects that cause deviations of the measured value from the true value.
The details in quantifying measurement uncertainty vary greatly depending on the measurement context and the application area. Below, we give a brief overview of the processes involved in quantification.
Evaluate the process
We can evaluate the measurement process to identify potential sources of uncertainty. Questions which we may ask ourself include:
- What measurement method will I use?
- What are my target test points?
- What is the range of my measurement capability?
The answers will help us identify sources of uncertainty such as instrument limitations, environmental conditions, human factors, calibration procedures, sample variability, etc.
Estimating uncertainty
There a number of ways we can estimate uncertainty but the most common way is through using the ISO Guide to the Expression of Uncertainty in Measurement (GUM) approach.
You can download the guide here: https://www.bipm.org/en/publications/guides/
Here is an overview of the GUM 8-step process for calculating uncertainty:
- Describe the Measurement Context: Begin by explaining the significance of the measured value within your specific measurement process.
- Identify Inputs and Uncertainties: List the input quantities and their associated uncertainties, including any potential correlations.
- Combine Uncertainties: Correctly combine the individual uncertainties to obtain the combined uncertainty.
- Apply Coverage Factor: Multiply the combined uncertainty by an appropriate coverage factor (usually a constant like 2 for a 95% confidence interval).
- Report Results: Present the final measured value with its uncertainty in the appropriate format, ensuring it aligns with your measurement process requirements.
By following these steps, you’ll ensure a thorough and accurate measurement process with clear documentation of your results.
If the measurement procedure is well-known then the important uncertainty sources are usually also known. You can find these in the relevant standard test method such as ISO standards and metrology institute websites, such as NMI (Australia) NIST (US), NPL (UK), BIPM (Global) also offer great resources.
Efforts should be made to minimise uncertainty sources by optimising the measurement procedure. The remaining uncertainty sources will then have to be taken into account in uncertainty estimation.
Example of sources of uncertainty in measurements includes:
- repeatability uncertainty
- calibration uncertainty
- environment uncertainty
- Application-specific uncertainty
Once the sources of uncertainty are identified, we can combine their individual contributions and obtain the estimate of measurement uncertainty, which in this case will be called combined measurement uncertainty.
You must research each factor in depth, looking at old data or conducting new experiments to quantify results.
The specific method for combining uncertainties depends on the nature of the sources and their relationships. You may also consider factors such as correlations between different sources in the calculations.
A great resource for this is Uncertainty of Measurement Results from NIST
Reporting on measurement uncertainty
Reporting of measurement uncertainty is dependent on its intended use but the basic principle of what is required remains unchanged:
When reporting the result of a measurement and its uncertainty, it is recommended that you err on the side of providing too much information rather than too little. For example, one should
a) describe clearly the methods used to calculate the measurement result and its uncertainty from the
experimental observations and input data;
b) list all uncertainty components and document fully how they were evaluated;
c) present the data analysis in such a way that each of its important steps can be readily followed and the
calculation of the reported result can be independently repeated if necessary;
d) give all corrections and constants used in the analysis and their sources.
A test of the foregoing list is to ask oneself “Have I provided enough information in a sufficiently clear manner that my result can be updated in the future if new information or data become available.
For instance, there are different guidelines for reporting measurement uncertainty values in calibration certifications and test results.
However, an overview of the process is given below.
First, you round uncertainty to two significant figures using conventional rounding methods. Then, you round the measurement result to match the uncertainty. For instance, if uncertainty is rounded to three decimal places, your measurement result should also round to the same.
You can then report the results alongside your test results, in a tabular format or as an English uncertainty statement. The statement reflects the range within which the actual value of the measured quantity is likely to lie. It is typically expressed as an interval with a specified confidence level, represented by a percentage value. A general statement template is given below.
“Reported uncertainties were estimated in accordance with [method name] expressed to a [value of confidence level] % confidence interval where k= [value of k].”
The uncertainty statement of the scale calibration should include sufficient information that a person not associated with the calibration will understand what uncertainty components were included, how and why.
If you are looking to learn more about this topic, we’ve found a great free course from MOOC: Estimation of measurement uncertainty in chemical analysis (analytical chemistry) course (ut.ee)
Measurement Uncertainties and Accreditation Standards
ISO/IEC 17025:2017 is the global laboratory competence standard recommended for all laboratories undertaking sampling, testing, or calibration work. Accreditation demonstrates reliability and increases trust among customers.
The standards outline the steps and principles laboratories should follow to quantify and report uncertainty effectively. They provide a common framework for laboratories worldwide, ensuring consistency and comparability in uncertainty evaluation.
One of the essential requirements of ISO 17025 is that laboratories demonstrate a well-established measurement uncertainty. Accreditation is awarded only to those laboratories that can:
- Demonstrate a comprehensive understanding of uncertainty sources that affect their measurements.
- Estimate uncertainties mathematically using appropriate statistical methods.
- Report their measurement results with associated uncertainty values.
Even though some guidance is given, laboratories must independently identify and mathematically compute uncertainties for their specific scenarios. SmartQMS is an Australian-based science and medical technology consulting firm that provides expert advisory and QMS development services to fast-track ISO certification for upcoming laboratories. Contact us today to get started.
