Type A evaluations

Type A evaluations use the statistical analysis of a series of observations (GUM document 2.3.2) and is based on practical measurements made in the laboratory, examples are those describing repeatability and reproducibility.

Type A evaluations do not investigate the cause of variation but quantifies the variation of an influencing factor on an end measurement result. Type A evaluations cannot ensure that all variations observed in the study are accounted for as they are limited to the period of time the measurements are being made.

Results of repeated measures are expressed as either the experimental standard deviation or experimental standard deviation of the mean (if several measurements are made).

Analysis is by standard statistical formulas commonly used in pathology laboratories:

The Standard Deviation (SD)

For most practises where the SD is assessed for laboratory assays >30 data points are used and therefore a normal distribution is observed. This means that there is no requirement for transformation of the SD – the standard uncertainty is simply equal to the SD. When <30 samples are tested, and normality cannot be assumed an alternative distribution such as the students t-distribution may be used. The application of the SD as we are using it here is to detect, and quantify, random, deterministic and stochastic errors. Systematic error is not assessed by the SD. Of note is that to be analytically useful variation of the process should be assessed over both the short and long term.

The Standard Deviation of the Mean

The standard deviation of the mean is used when several independent measurements are made and each has a mean and SD calculated. The SD of the means will then give the standard deviation of the mean.

Which of one of the SD or the Standard deviation of the mean is used depends on the nature of the result that is intended to be reported. For a single measurement, SD is used. For repeated measurements, from which a mean is calculated the SD of the mean is the preferred option.

All data used for Type A analysis should be readily available in the laboratory and there should be no need for additional testing.

Move on to Type B evaluations 

%d bloggers like this: