Sample size analysis for two-sample linear rank tests

Sample size analysis is a key part of the planning phase of any research. So far, however, limited literature focusses on sample size analysis methods for two-sample linear rank tests, although these methods have optimal properties at different distributions. This paper provides a new sample size analysis method for linear rank tests for location shift alternatives based on score generating functions. Results show a slightly anti-conservative behavior, no severe risk of an occuring circular argument at small to moderate variances of the population's distribution, and good performance compared to alternate sample size analysis methods for the most well-known linear rank test, the Wilcoxon-Mann-Whitney test.