[Estimation of reproducibility and repeatability in microbiological ring trials--robust versus conservative methods].

: During the last years there was a lot of progress to be seen in the development of standardized methods for microbiological ring trials. The statistical analyzing strategies, in particular the calculation of estimations for the parameters repeatability and reproducibility, will be considered in this paper. Apart from the conservative method of the variance analysis robust methods are increasingly discussed. We will compare and discuss these methods using data of recently realized ring trials. If we can assume a normal distribution of our data, then all estimations are theoretically precise and efficient. But up to now, we know very little about the character of the robust estimations, if the normal distribution cannot be assumed. In addition to this, we have to mention once more, that the use of robust estimators is unreasonable without taking a critical look on the data themselves. Thus, we will show the possibilities of graphical presentation of all data to identify laboratories with critical results.