Statistical analysis of change in ordered categorical classifications
Elisabeth Svensson
Mathematical Statistics
Chalmers and Göteborg University
Abstract
Subjective judgements of complex variables are commonly recorded in
ordered categorical data. The rank-invariant properties of such data
are well known and there are various statistical approaches for
analysis and modelling ordinal data. My presentation focus on the
property of non-additivity of ordered categorical data in the analysis
of change between two occasions.
A rank-invariant non-parametric approach that is valid regardless of
the number of response categories is presented. The method makes it
possible to separately measure the individual order-preserved
categorical changes that are attributed to the group change, and the
individual categorical changes that are not consistent to the pattern
of group change. Application examples from clinical rating scales with
the number of possible response categories ranging from three to one
hundred are given.