| 英文摘要 |
The recently developed Growth Mixture Mode (GMM) provides multiple trajectories to account for the heterogeneity of population, and is therefore more comprehensive than Latent Growth Curve Model (LGCM) that uses a single trajectory to describe development of all subjects, based on its homogeneity assumption. Nonetheless, without a strategic model development mechanism, researchers often encounter convergence problem with GMM, due to model complexity and flexibility. The aim of this study was to fulfill this deficiency by establishing a standardized step-by-step model development procedure. Results from empirical data showed that the six-step procedure improved the likelihood of model convergence significantly. Results from empirical analyses concluded three classes of developmental trajectory of depression among Taiwanese adolescents including stably low depression, named "cheerful" (82.3%); "start low end high", named "late onset depression"(7.7%); and "start high end low", named "early onset depression" (9.9%). GMM is a promising method with model-based longitudinal classification, and the mechanism proposed makes a significant contribution to GMM application. |
