### Improving uncertainty of linear calibration experiments

Improving uncertainty of linear calibration experiments

**Accuracy and Trueness – what is the difference?**

In analytical chemistry, we are quite familiar with the statistical term “accuracy” but of late, one may have noticed that the term “trueness” has been referred frequently, particularly in the evaluation of measurement uncertainty by the holistic top down approach which looks at the overall performance of the test method.

Is there a real difference between the meanings of these two terms?

ISO 5725-1 “*Accuracy (trueness and precision) of measurement methods and results — Part 1: General principles and definitions*” states the following definitions:

**Clause 3.6**

**accuracy **The closeness of agreement between a test result and the accepted reference value.

[SOURCE: ISO 3534-1]

**Clause 3.7 ****trueness**

The closeness of agreement between the average value obtained from a large series of test results and an accepted reference value.

[SOURCE: ISO 3534-1]

From these definitions, we see that “accuracy” is actually a qualitative term and as suggested by the title of ISO 5725-1, accuracy covers trueness and precision. On the other hand, “trueness“ is statistically quantifiable if we know the closeness of the mean values of repeated testing of a test material from its assigned or certified value. A high level of trueness is equivalent to a lack of bias in the test method.

We can assess “trueness” by one of the following manners:

- A significance Student’s t-test is applied to the null hypothesis that the method is free from bias when we measure the analyte content of a certified reference material several times;
- If such reference materials are not available, we can analyze a test portion of the sample before and after adding a known mass of analyte to it, a process known as “spiking”. If the difference between the two average measurements is not equivalent to the amount added, i.e. the recovery is significantly different from 100%, we conclude that some bias exists in the method.
- We can also compare the results from the test method which is being validated with the results obtained when a standard or reference method is applied to the same test materials. In this approach, we have to test a number of test materials containing different analyte levels. We can then apply the paired t-test or by regression methods to check for any significant difference between them.

**Celebrating a new milestone …..**

- ANOVA
- Design of Experiments
- Linear regression
- Significance testing
- Decision Rule
- Randomization
- Probability
- hypothesis testing
- outliers
- Sampling statistics
- Normal distribution
- Type I and II errors
- Sampling
- Sampling uncertainty
- Student's t-distribution
- Median
- Probability distribution
- Conformity testing
- Confidence interval
- GUM
- uncertainty
- Microbiology
- Decision rules
- Measurement error
- F-test
- Degrees of freedom
- Control chart
- Variance
- Risk
- profiency testing
- propagation of uncertainty
- p-Value
- Monte Carlo
- Anderson-Darling
- Precision
- ISO 17025
- Measurement bias
- t-test
- interlab comparison
- Bias
- standard uncertainty
- measurement uncertainty
- repeatability
- Confidence limits
- Detection limit
- Central limit theorem
- IQR
- Law of Averages
- Coverage factor
- Accuracy
- How to
- ISO FDIS 17025
- Method validation
- Excel spreadsheet
- Risk assessment
- Risk analysis
- Robust statistics
- Cross-checks
- Factorial design
- Chi-square
- quartile
- Divisor
- Aerobic plate count
- Compliance
- ISO 17025:2017
- Recovery
- Shewhart
- She
- Trueness
- Mean value
- analytical uncertainty
- standard additions method
- ISO/IEC 17025
- systematic error
- Internal quality control
- standard deviation

## Recent Comments