Date of Award

3-1993

Document Type

Thesis

Degree Name

Master of Science

Department

Department of Operational Sciences

First Advisor

Kenneth W. Bauer, Jr., PhD

Abstract

Factor analysis is a multivariate statistical procedure for analyzing and reducing large data sets. Many factor analysis schemes and techniques are available that lead to strikingly different results from the same data. This research effort used a Monte Carlo approach to investigate the properties of two rotation methods for simple structure, Kaiser's raw and normal varimax criterion. Data sets were developed from a set of contrived experimental factor patterns by multiplying each factor pattern by its transpose to create a covariance matrix. Data sets of multivariate normal deviates were in turn generated from each covariance matrix via the Choleski algorithm. Rotated factor pattern matrices from each data set were compared to their respective experimental factor pattern on the basis of structure, loadings and eigenvalues. These performance issues are addressed through regression analysis and separate factor analysis in which the grand mean of proposed measures of effectiveness are predicted. These measures of effectiveness include structure matching and root mean square error between the experimental and observed factor patterns. Several methods of characterizing factor pattern complexity and predicting rotation criterion performance are explored.

AFIT Designator

AFIT-GOR-ENS-93M-27

DTIC Accession Number

ADA262516

Comments

The author's Vita page is omitted.

Download

Share

COinS