종단패널 분석에서 측정동일성 확보와 비수렴 문제 해결 전략: 고령화연구패널(KLoSA) 적용 = Measurement Invariance and Non-Convergence Solutions in Longitudinal Panel Analysis: Application to the Korean Longitudinal Study of Aging

다국어 초록 (Multilingual Abstract)

This study longitudinally analyzed changes in life satisfaction among older adults using the Korean Longitudinal Study of Aging (KLoSA) data and explored methodological approaches to ensure measurement invariance and model stability. The analysis confirmed that while construct equivalence was maintained in the measurement of life satisfaction, partial invariance approaches were necessary due to differences in factor loadings and intercepts across time points. The initial model encountered issues of multicollinearity and non-convergence caused by specific response patterns, which were addressed through variable reduction, allowing covariances, and applying Bayesian estimation methods. The final model demonstrated excellent fit indices, indicating model stability. Latent growth modeling (LGM) revealed a gradual decline in life satisfaction over time in later life, with a more pronounced decrease observed in advanced old age. These changes were closely associated with health status, economic stability, and social relationship factors. This study highlights the importance of longitudinal analysis and model validation methods in research on life satisfaction in later life and provides foundational data for policy development in response to an aging society.