Abstract
This paper presents an approach leading to unique interpretations about reliability and validity in the confirmatory factor analysis model for dealing with multitrait-multimethod matrices. The correlated trait-correlated method (CT-CM) model does not necessarily identify the best solution, although it gives a unique interpretation for the data. On the other hand, the correlated trait-correlated method minus one (CT-C (M-1)) model (Eid, 2000) invariably identifies a solution, but the solution is not unique, depending on the arbitrary choice of which method factor is eliminated from the model. In this study, the constraint that the sum of the method factor scores equals zero is introduced into the CT-CM model, which then ensures a unique solution. These three models are applied to 12 actual correlation matrices, and a simulation study to verify the proposed model. The results indicate that the proposed model is more practical than the other conventional models.
