A comparison of principal component analysis, common factor analysis, and image analysis : a Monte Carlo study

Date

1993

Authors

Woodward, Todd Stephen

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

Principal component analysis, common factor analysis, and image analysis are compared on their ability to reproduce loading patterns underlying simulated data sets. Sample size, number of variables per factor, saturation, and percentage of measurement error were systematically varied. Common factor analysis was found to be more accurate at retrieving the population loading pattern when there was no measurement error, but principal component analysis gave the clearest representation of the population loading pattern when there was 25% measurement error or greater. Image analysis severely underestimated the loadings, except when there was no measurement error. The technique differences diminished as the number of variables per factor increased, as population. Loading saturation increased, and as the amount of measurement error decreased. Numerical explanations for these differences are presented, and recommendations for future research are proposed.

Description

Keywords

UN SDG 5: Gender Equality

Citation