성장혼합모형에서 잠재계층과 연속형 결과변인 간 관계 추정방법의 비교

Abstract

This study aims to provide researchers with guidelines on the appropriate estimation method to estimate the relationship between latent class and outcome variables in growth mixture models through Monte Carlo simulation. To this end, this study briefly introduces 1) the one-step method, 2) the PC method, 3) the Lanza method, 4) the traditional three-step method, 5) the new three-step method, and 6) the BCH method to estimate the relationship between latent class and outcome variables in growth mixture models, and compares the six estimation methods based on four performance criteria: relative bias, variance, power, and type I error through Monte Carlo simulation. The results of this study showed that the Lanza and BCH methods were superior across a range of conditions, with the Lanza and BCH methods being superior when the effect size was small and the Lanza method being superior when the effect size was large. Based on these findings, the significance, limitations, and follow-up studies were discussed.