Statistical Analysis of Complex Round Robin Tests

Different possibilities for the evaluation of complex round robin tests are presented. The advantages of 3-way over classical principal components analysis when applied to 3-way data arrays are discussed. The method allows extreme data reduction without essential loss of information. This is useful first of all for a graphically oriented evaluation of laboratories. The results of 4 round robin tests (8 laboratories, 5 parallel estimations, 4 trace element concentrations) for the analysis of heavy metals in waters (synthetic solutions with known concentrations) serve as an example. Special attention is paid to the recognition of different kinds of errors (random and systematic errors or unstable working).