On Modeling Missing Data in Structural Investigations Based on Tetrachoric Correlations With Free and Fixed Factor...

Area of Science:

• Psychometrics
• Statistical Modeling
• Data Analysis

Background:

• Missing data in statistical analyses can bias results.
• Confirmatory Factor Analysis (CFA) is a common technique for examining latent structures.
• The modeling missing data approach uses a latent variable to account for missingness.

Purpose of the Study:

• To extend the modeling missing data approach to utilize tetrachoric correlations as input.
• To investigate the impact of switching between free and fixed factor loadings in CFA models with missing data.
• To evaluate the performance of CFA models with and without a missing data latent variable.

Main Methods:

• A simulation study was conducted using confirmatory factor analysis (CFA).
• Models with and without a missing data latent variable were compared.
• Tetrachoric correlations were used as input, with variations in dataset size and missing data amount.
• Root Mean Square Error of Approximation (RMSEA) and Comparative Fit Index (CFI) were used for model fit evaluation.

Main Results:

• An additional missing data latent variable helped recover model fit when using tetrachoric correlations.
• The Comparative Fit Index (CFI) tended to overestimate model fit.
• Results aligned with modeling missing data assumptions when using fixed factor loadings.
• Partial agreement was observed for other conditions, indicating sensitivity to model specification.

Conclusions:

• Modeling missing data with a latent variable is effective when using tetrachoric correlations.
• Fixed factor loadings are recommended for the modeling missing data approach in CFA.
• Careful consideration of model specification is crucial when handling missing data in psychometric analyses.