scope:

General

An interlaboratory study aims to determine the comparative performance characteristics (trueness and precision characteristics) of the alternative method against the reference method.

Guidelines and requirements for organizing, dispatching and conducting the interlaboratory study are given in Annex H and in ISO 5725-2.

At least eight laboratories shall participate in an interlaboratory study.

The organizing laboratory is responsible for the preparation of the test protocol and a data sheet (see below) to be used by each laboratory for recording all measurement results and critical experimental conditions (see H.3).

The analyst in each collaborating laboratory shall demonstrate competence in the use of the alternative method and of the reference method prior to participating in the study.

In microbiology, the data {y} of repeated measurements do not always show a normal (Gaussian) distribution. Therefore, the distribution of these data should be checked for normality if more than 30 values are available at the same level. In order to get a more symmetric distribution, take logarithms of the counts.

The data obtained from interlaboratory studies often contain outliers, i.e. measurement values that deviate so much from comparable measurements that they are considered inconsistent. If they are retained in the data set, the trueness and precision characteristics (averages, standard deviations, etc.) obtained with classical methods of statistical analysis are unreliable. Therefore, ISO 5725-2 includes outlier tests (Cochran, Grubbs) in order to detect and eventually discard outliers and to obtain reliable trueness and precision characteristics. Often this causes disputes as to which of the outliers should be discarded from or retained in the statistical analysis. In order to avoid such disputes, robust estimates of the trueness and precision characteristics are used in this International Standard. Since they are insensitive to any extreme values, they always use the complete data set obtained from the interlaboratory study.

However, each extreme value should be checked for a clerical error in transcribing the measurement result, an error in computation, a slip in performing the measurement or an analysis of the wrong sample. If possible, such values should be replaced by the correct values. Other extreme measurement results or laboratories reporting extreme values are not excluded from the statistical analysis unless exclusion is based on sound microbiological reasons.