Development of Cluster Analysis Methods Suitable for Student Opinion Data.
Abstract
The Naval Postgraduate School's Student Opinion Form data were subjected to study through the use of two cluster analysis techniques: (1) K-MEANS partitioning method and (2) Chernoff's FACES. Much developmental work was performed to tailor these methods to the special requirements of the data set. A thorough multivariate statistical review provided the basis for choosing optimality criteria and distance functions for use in the MIKCA (Multivariate Iterative K-MEANS Clustering Algorithm). Alterations were made to the computer code to allow the analysis to include the effect of class size on cluster membership. Use of the linear discriminant function aided in identifying variables for use in constructing features of the computer-drawn faces. This approach to the Chernoff's FACES technique shows promise but needs further development. A principal components analysis of the data showed it to be essentially one dimensional. Partitioning the data into four clusters shows that the scoring of the courses varies inversely with class size. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1979
- Accession Number
- ADA068544
Entities
People
- Joel Weston Aiken
Organizations
- Naval Postgraduate School