Multiple Indicator Models

In multiple indicator models, there are several indicators for one trait–method unit. In the less restrictive model, there is one factor for all indicators belonging to the same trait–method unit. The correlations between these factors constitute a latent MTMM matrix. The correlation coefficients of this latent MTMM matrix are not distorted by measurement error and allow a more appropriate application of the Campbell and Fiske criteria for evaluating the MTMM matrix. Multiple indicator models allow the definition of traitspecific method factors and, therefore, the separation of measurement error and method-specific influences in a more appropriate way. Eid and colleagues have shown how different models of CFA can be defined for different types of methods. In the case of interchangeable methods, a multilevel CFA model can be applied that allows the specification of trait-specific method effects. In contrast to the extension of the CTCU model to multiple indicators, the multilevel approach has the advantage that the number of methods (e.g., raters) can differ between targets. In the case of structurally different raters, an extension of the CTC(M 1) model to multiple indicators can be applied. This model allows a researcher to test specific hypotheses about the generalizability of method effects across traits and methods. In the case of a combination of structurally different and interchangeable methods, a multilevel CTC(M 1) model would be appropriate.