Models for fuzzy nominal data

Numerical techniques and measurement procedures are proposed for processing nominal (or qualitative) data in situations where coders must assign observations or data to categories in an a priori classification scheme. Such situations arise most commonly in survey research, where data from open-ended questions must be rendered compatible with the requirements for computerized data sets. These techniques have the advantage of allowing the coder to make categories flexible rather than hard-edged. They also provide statistical guidelines for assessing the relative empirical worth of researchers' category schemes. The framework within which the techniques are developed is based on the theory of fuzzy sets. The fundamental concept involved is that an element may belong partially to a set rather than belonging either totally or not at all, and an element may belong partially to more than one set simultaneously. Topics covered in the paper include: P ]

1. Measuring the degree of coder ambiguity about category membership of data;
2. Evaluating the fuzziness of category schemes;
3. Diagnosing the sources of fuzziness in such schemes; and
4. The multivariate analysis of fuzzy nominal data.