The Performance of Growth Mixture Modeling With Skew-T Distribution and Covariates

• In Event: 46.079 - Appropriate Modeling of Latent Growth and Relationships

Abstract

Class determination is an unresolved problem in growth mixture modeling (GMM). As Muthén and his colleagues proposed and implemented in the most recent version of Mplus, GMM with skew-t distribution may relax the assumption of with-class normality of GMM and so the problem of spurious latent class caused by non-normal distributed variables may be avoid. The current Monte Carlo study aims to investigate the performance of a GMM with skew-t distribution in identifying the number of latent class and estimating parameter and standard error when different degrees of nonnormality are encountered in data. Moreover, the effects of covariates are also examined to confirm previous studies. Based on simulation results, this novel model is applied to a real data.

Author

• Min Liu, University of Hawaii - Manoa