Combined Approach to Multi-Informant Data Using Latent Factors and Latent Classes: Trifactor Mixture Model

Area of Science:

• Psychometrics
• Quantitative Psychology
• Educational Measurement

Background:

• Collecting data from multiple informants is crucial but lacks robust statistical methods for modeling informant congruence and incongruence.
• Existing trifactor models decompose variance but do not simultaneously address target heterogeneity.
• The trifactor model by Bauer et al. separates common, informant-specific, and item-specific variances.

Purpose of the Study:

• To introduce and demonstrate the trifactor mixture model, integrating trifactor and mixture models.
• To provide a statistical framework for analyzing common and unique perspectives from multiple informants while accounting for target heterogeneity.
• To illustrate the model's application using student self-rated and teacher-rated academic behaviors.

Main Methods:

• Development and application of the trifactor mixture model.
• Utilized a large dataset of student self- and teacher-rated academic behaviors (N = 24,094).
• Detailed explication of model specification, testing procedures, and practical considerations for trifactor mixture analysis.

Main Results:

• The trifactor mixture model successfully integrates informant perspectives and latent class analysis.
• Demonstrated the model's capability to differentiate shared variance, informant-specific variance, and item-specific variance.
• Showcased the identification of latent classes representing target heterogeneity in academic behaviors.

Conclusions:

• The trifactor mixture model offers a powerful statistical tool for multi-informant research.
• This approach enhances the nuanced understanding of agreement and disagreement in behavioral data.
• Facilitates more sophisticated analysis of psychological constructs across different perspectives and target groups.