[Use of nonparametric methods in medicine. IV. Comparison of levels in more than 2 samples].

The authors mention two tests for comparison of the level of random variables in more than two random samples. It is thus a question of generalization of methods presented in parts II and III. Friedman's test resolves the position for dependent samples when data assembled during different experimental situations in the same group are to be evaluated. This test calls for a complete series of measurements from every statistical unit, in mathematical terminology this means that the samples are equally extensive. In case of large samples Friedman's test is based on chi 2 distribution; for a small number of samples special tables of critical values are given. Kruskal-Wallis's test resolves the position for independent samples. It is also based on chi 2 distribution.