Internal Multi-Dimensional Scaling of Categorical Variables
Abstract
The purpose of the study in the dissertation is to translate raw categorized data into numerical values on which standard statistical analyses can be performed. When raw observations are recorded on a nominal scale, they are to be transformed so that the resulting numbers can be regarded as lying on an interval scale. A scalling technique is developed on the basis of a generalization of Lancaster's approach (canonical correlation for two sets). The report also presents computer programs starting from data in contingency tables which are converted into a correlation matrix. Initial values are used in order to start the minimum-determinant process. Various initial weights and the final minimum-determinant solution are compared.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1974
- Accession Number
- AD0782706
Entities
People
- Jeffrey Chit-fu Chang
- Rolf E. Bargmann
Organizations
- University of Georgia