skip to main content skip to footer

Gauss-Seidel Computing Procedures for a Family of Factor Analytic Solutions

Author(s):
Browne, Michael W.
Publication Year:
1968
Report Number:
RB-68-61
Source:
ETS Research Bulletin
Document Type:
Report
Page Count:
74
Subject/Key Words:
National Institute for Personnel Research, Computation, Factor Analysis, Least Squares Statistics, Maximum Likelihood Statistics, Statistics

Abstract

Gauss-Seidel computing procedures have previously been used to obtain Maximum Likelihood and Minres factor analytic solutions. This paper is concerned with the use of Gauss-Seidel computing procedures to obtain five solutions: Maximum Likelihood, Minres, Alpha and two recently derived weighted least squares solutions. The treatment of negative residual variances is considered, modifications of approaches suggested by Harman and Fukuda (1966) and by Joreskog (1967a) being used. Three "generalized characteristic value" problems occur and procedures for solving them are provided. Finally, examples of the application of the Gauss-Seidel procedures are given and are used to illustrate properties of the five factor analytic solutions.

Read More