B0420
Title: Ranking procedures for repeated measures design with missing data
Authors: Kerstin Rubarth - Institute of Biometry and Clinical Epidemiology - Charite Berlin (Germany) [presenting]
Frank Konietschke - Charite Berlin (Germany)
Abstract: A frequently used design in health, medical and biomedical research is the repeated measures design, where units, e.g. patients, are observed several times under different conditions. However, if data is highly skewed, ordinal or even ordered categorical or if sample sizes are small, only a few methods are applicable as many methods, e.g. linear mixed models, rely on the assumption of multivariate normality and a specific covariance matrix structure of the error terms. Additionally, in many studies with repeated measures, missing values are almost certain to occur. Therefore, we propose a purely nonparametric method for the analysis of repeated measures with missing data. The hypotheses are formulated in terms of the nonparametric treatment effect. Global testing and a multiple contrast test procedure, as well as simultaneous confidence intervals, are developed for this design. We present simulation studies, which indicate a good performance of this procedure in terms of the type-I-error rate and the power under different alternatives with a missing rate up to 30\%, also under non-normality. A real data example illustrates the application of the newly developed methodology.
