ABSTRACT
Almost all data sets we have investigated show clustering of judges to some extent. The Political Goals data has the Materialist/Postmaterialist dichotomy; the Living Places data splits into groups based on the degree of urbanness the judges enjoy; Occupations are Managerial or Technical; there are Team and Non-team Sports. In many cases, the experiment is set up to reveal groupings, and other times they arise naturally. Some models are more adept at accommodating such grouping than others. The distance models, Plackett-Luce, and Bradley-Terry /Mallows models presume a certain degree of unimodality in the judges' rankings. More complicated models, such as the (extended) Marginals and Babington Smith models, and the orthogonal contrast tjJ and Free models, are flexible enough to capture additional dimensionality. By the same token, they are somewhat less simple in their interpretation.
