Albers, Sönke and Lutz Hildebrandt
Methodische Probleme bei der Erfolgsfaktorenforschung - Messfehler, formative versus reflektive Indikatoren und die Wahl des Strukturgleichungs-Modells
Schmalenbachs Zeitschrift für betriebswirtschaftliche Forschung, 58 (1): 33-2, (2006).
Abstract: We demonstrate methodological problems with the application of the so called “Cronbach’s α — LISREL”-paradigm. Working with unobserved latent constructs operationalized through indicators allows for differentiating into a measurement and structural model. However, while the elimination of internally inconsistent indicators is correct for reflective indicators, this proves to be incorrect for the mostly formative indicators in success factor studies. This neglects important facets of the problem which lead to results that vary according to the included indicator set. In this context, the analysis of structural equation models with LISREL is not appropriate because it can handle only reflective indicators while PLS allows for formative indicators. However, in the latter case it is often better to apply ordinary least squares regression to the battery of indicators because multicollinearity has to be dealt with in both, PLS and OLS, while regression models can handle nonlinear relationships as well as unobserved heterogeneity.Show all publications