Exploring and testing cluster analysis |
| |
| Authors: | Bruce A Peseau Roger L Tudor |
| |
| Institution: | (1) Program in Administration of Higher Education, The University of Alabama, P.O. Box Q, 35487-9782 Tuscaloosa, AL;(2) Department of Statistics and Computer Sciences, Gadsden State Community College, Gadsden, AL |
| |
| Abstract: | Academic deans annually negotiate for a fair share of university resources. Without quantitative accreditation standards to serve as minimum guidelines, the competition is more difficult. Peer institution comparison, derived from the analysis of quantitative data on resource and productivity variables, assists by providing less biased justification for budget requests. In this study, data from a national study of doctoral-granting teacher education programs was used for a factor analysis of key resource and productivity variables. Seven alternative cluster analysis methods were used and compared. Three follow-up procedures (analysis of variance, discriminant analysis, and classification analysis) tested the cluster analysis results to verify the differences between clusters and their internal homogeneity on three different cluster solutions. The classification analysis showed that 88.2% of the 76 programs were correctly classified on the five-cluster solution produced using Ward's minimum variance method. The practical use of such data suggests that Cattell's pattern of profile similarly is a useful beginning to compare programs on factor scores; following that, raw data on each key variable might be used to prepare individual program profiles for more specific comparison. |
| |
| Keywords: | |
| 本文献已被 SpringerLink 等数据库收录！ |
|
