Contributions to generalized Wilcoxon rank tests

Abstract: In unbalanced small sample size problems with right censorings, the variance estmators of linear rank statistics, may be biased. This is the case with the commonly applied variance estimators for the generalized Wilcoxon rank test statistics of Gehan and Prentice. The bias and variance of different variance estimators and observed significance level and power for standardized tests are compared in a Monte Carlo simulation study. The variance estimators are the permutational-, the conditional permutational- and the jackknife variance estimators for the statistic Gehan, and the asymptotic- and the jackknife variance estimators for the statistic of Prentice. It appears that observed level and power and variance properties may be improved by using the jackknife variance estimator.Further, The sensitivity curves of Turkey are used to establish the sensitivity to gross errors and misclassifications for standardized generalized Wilcoxon rank sum statistics in small samples with right censorings. For a certain combined sample, which might contain gross errors, a relatively fast method is needed to establish the applicability of the inference drawn from the selected ranktest. Ome way is to use the "change of decision point" (cdp), which in this thesis is defined as the smallest proportion of altered positions resulting in an opposite decision.