Control Charts and Measurement Uncertainty

Control charts are widely used in laboratories to monitor the drift of an instrument or measuring device, but their use generally stops there, even though these data can also be used to estimate the laboratory’s measurement uncertainty.

To determine the expanded measurement uncertainty U from data obtained in the laboratory itself, two factors are important:
- trueness or bias: this may be considered the systematic error, whereas precision represents the random error associated with a measurement;
- within-laboratory reproducibility.

The bias can be further divided into two components: the bias value, b, and the uncertainty associated with: b, and the uncertainty associated with b: ubiais.

In addition to these two components related to bias, it is also necessary to take into account uRw, the uncertainty associated with the analytical process itself, as characterized by within-laboratory reproducibility.

The expanded measurement uncertainty U is calculated from these three components: one systematic component and two random components.

avec K = 2.

The following approach for determining measurement uncertainty is carried out as follows:
- determine the intra-reproducibility;
- determine the bias and its associated dispersion;
- calculate the total measurement uncertainty by combining all contributing factors.

The easiest approach is to work with relative values in order to combine uncertainties without difficulty.
To convert the standard deviation of a series of values into a relative standard deviation, the following formula should be applied: S% = 100 × s / x̄ where:
s = standard deviation of the series
x̄ = expected value or mean of the series

For example, on the control chart, with a mean of 103.6 ppm and a standard deviation of 2 ppm, then: S% = 1.9%

Relative within-laboratory reproducibility uRw% is a measure of the long-term dispersion of analytical results within a single laboratory, under varying conditions such as different operators, different standards, different times, and different environmental conditions.

Control chart data obtained from a CRM (Certified Reference Material) representative of the samples analyzed:

The relative variance of within-laboratory reproducibility is therefore equal to S², the relative variance determined from the CRM control chart or the control sample chart : u

Rw% = S%²

Control chart data obtained from a CRM (Certified Reference Material) that is not representative of the analyzed samples.:

Intermediate precision calculated from the control chart of a CRM and supplemented with data from determinations performed on routine samples.

Based on the control chart, determine uRw² = S%², The relative variance of the CRM,

Calculate the relative range d% (the relative differences for each routine sample):

where x1 et x2 represent the individual measurement values of the determination, and X represents their average.

Calculate the average relative range Rmoy%, the mean relative difference of the d% values,

Calculate based on Rmoy%, the relative standard deviation of repeatability RS% : RS% = Rmoy%/1,128 (estimating the experimental standard deviation according to FD X 07-021 from two values).

The relative variance of the measurement uncertainty due to intra-reproducibility is therefore obtained by combining the two variances in quadrature :

uRW%² = uRc²+RS%²

Control chart using a Certified Reference Material (CRM) that is not representative of the samples (X = 100 ppm and u = 2 ppm --> uRc = 2%)

Sample values:

X1 X2 X̄ |(X1-X2)| d%
102 108 105 6 5,7
102 102 102 0 0,0
104 106 105 2 1,9
102 108 105 6 5,7
98 102 100 4 4,0
104 106 105 2 1,9
Mean relative range Rmoy% 3,2

 

RS% = 3,2/1,128=2,8%

Determination of Bias Using a Control Chart for a Certified Reference Material (CRM)

Bias: systematic component of measurement error, corresponding to the difference between the expected value of measurement results and a reference value.

A CRM is characterized by two parameters : Cref,the certified reference value and UCref, the expanded measurement uncertainty associated with this reference value.

The reference value is used to determine the bias value, UCref must be taken into account when determining the uncertainty associated with bias.

The procedure is as follows :

- Calculate using the given expanded measurement uncertainty UCref, the relative measurement uncertainty uCref% using the following formula :

with k, the coverage factor= 2

Perform at least 6 analyses — the minimum required to reasonably expect a normal distribution — on the CRM, and calculate the mean measurement value from the analytical results x̄ and the standard deviation sMRC, and the relative standard deviation (RSD) RSMRC%,

- Calculate the relative bias b% from the certified value Cref and the mean measured value X using the following formula::

Calculate the relative standard deviation of the bias using the formula RSbiais = RSMRC/√n where n is the number of values used to calculate the mean x̄.

The uncertainty associated with the bias is given by the following root mean square:

If the relative bias is not significant.

Otherwise, it should not be included in the formula and should be corrected directly on the value.

Exemple:

100 ppm CRM with a standard uncertainty of 2 ppm.

uCref% = 2,00%

Six analyses were performed on this CRM: average x̄ = 103,6 ppm with a standard deviation of 2,16 ppm

RSMRC% = 2,16% et RSbiais = 2,16/√6 = 0,88%

b% = 3,6%

ubiais% = 4,2%

Expanded Uncertainty Calculation:

U = 2√(4,2² + 3,50²) = 11 % avec K = 2

X = (100±11) ppm