Conceptualizing Correlated Residuals as Item-Level Method Effects in Confirmatory Factor Analysis.

Area of Science:

• Psychometrics
• Statistical Modeling

Background:

• Correlated residuals in confirmatory factor analysis (CFA) can violate item uniqueness assumptions.
• Two-variable disturbances often hinder good model fit in CFA.

Purpose of the Study:

• To investigate representing item-level method effects as method factors in bifactor models.
• To assess model identification and efficiency of this approach compared to correlated residuals.

Main Methods:

• Conceptualized item-level method effects as method factors within a bifactor measurement model.
• Integrated method factors into fixed-links, parallel-measurement, or tau-equivalent submodels.
• Conducted a simulation study to compare submodel efficiency in controlling disturbances.

Main Results:

• Integrating method factors into fixed-links, parallel-measurement, or tau-equivalent submodels proved effective.
• All investigated submodels showed similar efficiency in managing two-variable disturbances.
• The fixed-links model and the correlated residuals model demonstrated perfect correspondence in fit results.

Conclusions:

• Modeling item-level method effects as method factors in bifactor models is a viable alternative to correlated residuals.
• Fixed-links, parallel-measurement, and tau-equivalent submodels offer efficient ways to implement method factors.
• This approach maintains the principle of unique residual variation for each item.