[What is the therapeutic effect if a patient gets better but nobody knows how much? Analysis of change when the data material consists of ordered categories].
Elisabeth Svensson · PubMed · 2005
Assessments on scales and questionnaires generate ordered categorical data. Characteristic feature is the ordered structure with lack of information regarding size and inter-categorical distances. These limited mathematical properties of ordered categorical data mean that the most common statistical methods for evaluation of change, the Student's paired t-test and the Wilcoxon sign rank test, are not appropriate to use as they are based on differences. This paper presents classical methods useful for dichotomous data but they are appropriate for dichotomised ordered categorical data as well.