Statistical methods for paired ordinal data

Questionnaires and rating scales are used in most disciplines. Irrespective of the application and the type of scaling, assessments on scales generate ordinal data having rank-invariant properties only, which means that non-parametric statistical methods must be used for description and analysis. To this end a family of statistical methods based on augmented ranking procedures has been developed in the project. This allows the systematic part of an observed disagreement in paired assessment to be identified and measured separately from additional individual variability. Important applications are inter- and intra-reliability studies of scale assessments, evaluation of the levels of population-based and individual-based changes in ordinal data, studies regarding inter-scale consistency, validity and rank-invariant calibration. The methods are still under evaluation, especially regarding the distributional properties of the measures and development of statistical tests.