DESIGN AND EVALUATION OF THE QUASIMEME INTER-LABORATORY PERFORMANCE STUDIES - A TEST-CASE FOR ROBUST STATISTICS

The QUASIMEME laboratory performance studies have been designed according to the recent IUPAC/ISO protocol. An overview of the design is given and, in addition to the Z-score for ranking the bias of laboratories, a term called 'P score' is introduced as an aid to the assessment of the precision of the laboratories. The data have been evaluated using robust statistics. Little information is available on practical applications of robust statistics for the evaluation of inter-laboratory studies, therefore, part of the data have been used to compare two different algorithms for robust statistics and the conventional ISO 5725 procedure. The data obtained from these intercomparison studies are heterogeneous in nature, and it is concluded that robust statistics have distinct advantages in assessing the data, but should not be used without a critical examination of the dataset.