[Use of nonparametric methods in medicine. II. Comparison of levels in 2 related samples]

The authors present two non-parametric tests which are suitable for comparison of the level of a random variable on the basis of results obtained from two dependent samples. The sign test is discussed which is based on the evaluation of the statistical significance of the difference in the number of positive and negative deviations. The application of this test in practice is facilitated by the enclosed table which gives critical frequencies which still lead to the refusal of the zero hypothesis. The authors discuss also Wilcoxon's test for paired values. This test takes into account not only the trend of deviations but also their magnitude and thus uses more information than the sign test and has thus a greater scope. For this test an auxiliary table of critical values is also presented.