Multiple Imputation of Missing Data in Moderated Factor Analysis

Area of Science:

• Psychometrics
• Statistical modeling
• Data analysis

Background:

• Moderated factor analysis extends the common factor model with a continuous moderator variable.
• Handling missing data on indicator variables is typically managed with full information maximum likelihood.
• Missing data on the moderator variable presents a significant challenge, often necessitating listwise deletion.

Purpose of the Study:

• To propose and evaluate a multiple imputation procedure for moderated factor analysis with missing moderator data.
• To compare the performance of the proposed method against listwise deletion and predictive mean matching.

Main Methods:

• Development of a moderated factor model-based multiple imputation technique.
• Comparative analysis using simulated data under various missing data conditions.
• Evaluation metrics included parameter estimate bias and statistical power.

Main Results:

• The proposed multiple imputation procedure effectively handles missing moderator data.
• Listwise deletion and predictive mean matching exhibited lower power and greater bias in parameter estimates.
• Multiple imputation demonstrated superior performance compared to traditional methods.

Conclusions:

• Multiple imputation is a robust and recommended approach for addressing missing moderator data in moderated factor analysis.
• The proposed method offers improved accuracy and statistical power over listwise deletion and predictive mean matching.
• This advancement facilitates more reliable analyses in the presence of complex missing data patterns.