Confirmatory Factor Analysis
Subsequently, a confirmatory factor analysis (CFA) was carried out. This is justified by the increasing application of structural equation modeling (SEM) techniques, which have become standard in the assessment of multi-item measures over the years (e.g., Anderson & Gerbing, 1982; Martens, 2005; Martens & Hasse, 2006; Steenkamp & Trijp, 1997). CFA differs from EFA in that the former can be used to confirm a-priori hypotheses about the relationships on the diverse levels of the construct (i.e., the measurement model). EFA lacks this ability. Hence, CFA is a valuable instrument for testing the construct’s dimensionality or structure and can further be used to assess the reliability as well as validity of the construct (second-order), factor (first-order), and individual item level. The equally sized (n=425) developmental and the holdout sample were both processed independently by means of LISREL 8.50 for Windows (Joreskog & Sorbom, 1993). LISREL is a software package that is associated with the covariance structure analysis (CSA). A CSA approach was preferred to partial least squares (PLS) analysis, as it enables the estimation of measurement models, including error terms, the empirically comparison of model specifications and the assessment of the overall fit of (alternative) model structures. In addition, LISREL offers a number of diagnostics which can be helpful for model respecification (Diamantopoulos & Winklhofer, 2001). All confirmatory factor models discussed in this thesis have used a covariance matrix as input and the maximum likelihood estimation method.