Abstract
Although parametric methods are commonly used for longitudinal data, the use of these methods depends on several assumptions. In recent years, nonparametric methods have been developed to investigate the effect of multiple factors for the longitudinal data that do not meet parametric test assumptions and/or that include ordinal variables. The aim of this study is to introduce the nonparametric method for the analysis of longitudinal data which was developed by Brunner and colleagues and to implement it on a data set from the field of physical medicine and rehabilitation. Marginal distributions are used for obtaining parameter estimates and hypothesis testing, in the analysis of longitudinal data with nonparametric methods. While ANOVA or Wald type statistics are used to test the hypotheses depending on the sample size, relative treatment/marginal effects are used
as descriptive statistics. For the application, Duruöz Hand Index (DHl) scores are compared between two independent groups (control and treatment) at three time points; pre-treatment, during treatment and after 3 months of treatment, by using the FI-LD-FI design. The method developed by Brunner and colleagues offers solutions to different trial designs which may arise in the analyses of longitudinal data. Not requiring any test assumptions and any constraints in terms of sample size can be summarized as the advantages of this method.