Data-Generating Mechanisms Versus Constructively-Defined Latent Variables in Multitrait-Multimethod Analysis: A...

Area of Science:

• Psychometrics
• Quantitative Psychology
• Statistical Modeling

Background:

• Multitrait-multimethod (MTMM) studies are crucial for understanding construct validity.
• Castro-Schilo et al. (2013) evaluated models for relating MTMM data to external variables.
• Discrepancies were noted when fitting certain models to data generated by others.

Purpose of the Study:

• To question the primacy of the Correlated Traits-Correlated Methods (CT-CM) model in MTMM research.
• To propose the Correlated Traits-Correlated (Methods - 1) [CT-C(M - 1)] model as a plausible alternative.
• To advocate for Confirmatory Factor Analysis-MTMM (CFA-MTMM) models grounded in psychometric theory.

Main Methods:

• Re-parameterization of a basic MTMM true score model.
• Demonstration of the CT-C(M - 1) model's meaningful and parsimonious representation.
• Advocacy for CFA-MTMM models with explicitly defined latent variables.

Main Results:

• The CT-C(M - 1) model can be formulated as a reparameterization of a standard MTMM true score model.
• This re-parameterization results in a parsimonious and theoretically meaningful MTMM data representation.
• The CT-CM model's superiority as a data-generating model is questioned.

Conclusions:

• The CT-C(M - 1) model offers a viable and theoretically sound approach to MTMM data analysis.
• Explicitly defined latent variables in CFA-MTMM models enhance psychometric understanding.
• The choice of MTMM model significantly impacts the interpretation of associations with external variables.