ABSTRACT
In situations when hypotheses testing is the main focus of analysis and the design is balanced, that is, individuals are evaluated at the same time points, a non-parametric approach that does not assume a particular shape or form for the outcome and structure of the variances and covariances is more appropriate. This chapter mentions some classical non-parametric procedures in order to introduce non-parametric methods and outlines their advantages and disadvantages. It discusses the first non-parametric test used for repeated measures data, namely Friedman's test. The chapter emphases on the general approach for longitudinal data in factorial experiments presented at a non-technical level in the chapter. A commonly used but potentially misleading approach to the non-parametric analysis of repeatedly measured data is to rank all observations in the data set and then perform mixed model analysis on the ranked data, as usual.
