TABLE OF CONTENTS The statistical methods described in this chapter are the tools that allow the CRM user to compare a measurement result(s) to a certified value on the CRM certificate. In order to do this effectively there are a number of rules that need to be kept in mind. More detailed information on statistics in general can be found in reference [1].
Independent replication requires that all the stages of the analytical process are carried out separately in order to obtain individual results, i.e. a replicate result should not be influenced by previous common activities. For example, independent replication could mean that a com letely new instrument calibration is required for each replicate determination in order to remove the effect of a common calibration.
Whenever repeated analytical measurements are made there is always some variation in the results caused by, for example, changes in environmental conditions or a slight variation in the way the sample is introduced into an instrument. If the data are rounded before carrying out any statistical tests, the natural variation within the data set can be masked (artificially increased or decreased). As a result, the statistical tests may lose power or not work at all, e.g.
| 7.172, 7.175, 7.165, 7.183, 7.18 | Mean = 7.175 | Standard deviation = 0.007 |
| 7.2, 7.2, 7.2, 7.2, 7.2 | Mean = 7.2 |
