Classifying classification problems

Gower, John (2004). Classifying classification problems. In: Comptes Rendus des Iles Rencontres de la Societe Francophone de Classification (Chavent, M; Dordan, O; Lacomblez, C.C.; Langlais, M and Patouille, B eds.), Information not provided, pp. 41–47.



Classification problems, both for assignment and class construction, are specified either in probabilistic form or not. Underlying issues are (i) the types of sampling unit under consideration: in particular, are they differentiated into previously determined classes (possibly with identical members) or are they undifferentiated and (ii) considerations of the types of variable used; are they quantitative or categorical? Rather than a simple data-matrix, the fundamental form of data is taken to be the between and within-group structure. These considerations lead to a simple cross-classification of familiar, and some novel, classification problems.

Viewing alternatives

Item Actions