Reliability and latent class analysis as alternatives to factor analysis

Abstract

In this study the objective was to show that for categorical data, factor analysis might not be the right approach, to determine construct validity and to extract hidden factors. The dissertation is divided in three parts. First, an introduction to factor analysis is discussed, together with applications, on two different sets of categorical data, one for construct validity, and one for factor extraction. The dissertation continues on theory of reliability, as an alternative application to factor analysis, using Cronbach's alpha. An alternative to factor extraction is latent class analysis, which is next tackled in the dissertation, both theoretically and practically. Finally, it is concluded that factor analysis is not the right tool for categorical data.