The question of whether disorders are best described by underlying dimensions or by categorically distinct groups can be addressed by a statistical analysis of observed data of symptoms or questionnaire items. There are different methods that can be used for this purpose, and the three most commonly used approaches differ considerably with respect to their assumptions. The assumptions translate directly to the knowledge about the data that is required for a proper application of a given method. The main conclusion of this review is that the choice of method should match the required knowledge, and that it is advantageous to choose a method that permits integration of the existing knowledge in the analysis. Importantly, conclusions drawn from a particular analysis depend on the type of sample and the observed items, and should therefore be contextualized appropriately.
