[Falsely positive values in multi-channel analysis: an estimation of their frequency with Sylvester's formula and their reduction by a multivariate test quantity (author's transl)].

The application of the binomial distribution for determining the number of "falsely positive values" is not suitable because of the partly strong correlations between the clinical chemical parameters. We have therefore tried to obtain more precise estimations for the number of falsely positive values by using a generalisation from Sylvester's formula and taking into account the correlations. We use multivariate statistics for testing a patient's value" when several laboratory parameters are simultaneously considered. The multivariate technique leads firstly to a significant reduction of the erroneously interpreted "falsely positive values" and secondly allows the detection of hidden implausible constellations of values.