Study Length, Change Process Separability, Parameter Estimation, and Model Evaluation in Hybrid Autoregressive-Latent...

Area of Science:

• * Longitudinal data analysis
• * Structural equation modeling
• * Developmental psychology

Background:

• * Hybrid autoregressive-latent growth models synthesize autoregressive and latent growth frameworks for longitudinal data.
• * These models may face challenges in distinguishing between autoregressive and growth processes during estimation, potentially leading to biased parameter estimates.
• * Increasing the number of time points in a model can offer more information about change, potentially improving the separation of processes and parameter accuracy.

Purpose of the Study:

• * To investigate the relationship between change process separability, the number of time points, and model misspecification consequences.
• * To compare the performance of three prominent hybrid autoregressive-latent growth models: Latent Change Score (LCS), Autoregressive Latent Trajectory (ALT), and Latent Growth Model with Structured Residuals (LGM-SR).

Main Methods:

• * Monte Carlo simulation methods were employed to examine model performance.
• * The study focused on three specific hybrid autoregressive-latent growth models.
• * Key metrics included change process separability and robustness to misspecification across varying numbers of time points.

Main Results:

• * More time points enhanced process separability and robustness for LCS and ALT models, but often impractically so for researchers.
• * The LGM-SR model consistently exhibited high process separability and robustness to misspecification, irrespective of the number of time points.
• * Simulation results indicated that the LGM-SR model was the most effective among the evaluated hybrid models.

Conclusions:

• * The Latent Growth Model with Structured Residuals (LGM-SR) is recommended for its superior performance in analyzing longitudinal data using hybrid autoregressive-latent growth models.
• * Researchers should consider the LGM-SR model for more accurate and robust estimation of change processes in longitudinal studies.
• * While increasing time points can help, the inherent structure of the LGM-SR offers a more reliable solution for process separability.