Right-sizing growth mixture models as multi-group growth and confirmatory factor models

Area of Science:

• * Statistics
• * Quantitative Psychology
• * Developmental Science

Background:

• * Multi-group growth curve models with varying factor structures across classes form the basis for growth mixture models.
• * These models are crucial for identifying qualitatively different patterns of growth or decline within distinct classes.
• * Prior assessment of growth factor loading dimensionality and patterning is essential before determining the functional form of growth.

Purpose of the Study:

• * To introduce and illustrate the estimation of loading variant mixture models.
• * To compare candidate models and assess the performance of various fit indices under different sample sizes.
• * To demonstrate the application of these models in analyzing real-world data for superior model fit and conceptually distinct latent classes.

Main Methods:

• * Employed simulated datasets to estimate loading variant mixture models.
• * Compared candidate models using various statistical fit indices.
• * Analyzed a real-world dataset using a two-factor growth model approach.

Main Results:

• * Initial assessment of factor loadings prevents the identification of artifactual latent classes.
• * Loading variant mixture models offer a robust approach to identifying distinct growth patterns.
• * A two-factor growth model demonstrated superior fit and conceptual clarity over linear or quadratic mixture models in a real-world dataset.

Conclusions:

• * Loading variant mixture models are effective for identifying qualitatively different growth trajectories.
• * Careful examination of factor loadings is critical for accurate latent class identification.
• * The proposed methodology provides a more accurate and conceptually meaningful analysis of developmental patterns compared to traditional mixture models.