A test procedure based on ranks for the statistical evaluation of toxicological studies

For testing the significance of dose-related effects, a test procedure was developed for the typical situation in toxicological studies where several dosage groups are compared to one control group. Using a closed testing procedure in the sense of Marcus et al. (1976), the algorithm is sensitive to the detection of toxicological effects without increase in type-one errors. Furthermore, all significance tests applied in the algorithm are based on ranks and the calculation of the error probability is performed by a random-permutation technique (Wall 1974, 1985); the algorithm is therefore also applicable to non-normally distributed data. In the case of discrete quantities, this technique takes into account tie-structures of all kinds (Lüdin 1985). An application of the proposed test-procedure is given in the work of Kistler and Hummler (1985).