A nonparametric estimator of the shift effect for repeated observations

In a nonparametric two-sample model for independent observations with repeated measurements, a new point estimator and a new distribution-free confidence interval for the difference in means are introduced. The method is based on some ideas in Hodges and Lehmann (1963, Annals of Mathematical Statistics 34, 598-611). The asymptotic theory in Brunner and Neumann (1983, Biometrical Journal 24, 373-389; 1986; Biometrical Journal 28, 394-402) and the results for small sample sizes in Brunner and Compagnone (1988, Statistical Software Newsletter 14, 36-42) are used. The new estimators are applied to a problem in morphometry.