A Multivariate Latent Class Profile Analysis with Latent Group

EasyChair Preprint 874

Abstract

This paper suggests a new type of latent variable model which discovers the association between several categorical latent variables. A set of repeatedly measured categorical response variables forms a latent profile variable, while the other set of item variables identifies a latent group variable. Latent class profile analysis with group variable (GLCPA) explains an association between these two categorical latent variables as a form of two-dimensional contingency table. We applied GLCPA model to the NLSY 97 data to investigate the association between of depression process and the longitudinal behaviors of substance use development among adolescents who experienced an Authoritarian parental styles in their youth.

Keyphrases: latent class analysis, logistic regression, longitudinal data, recursive EM algorithm

@booklet{EasyChair:874,
  author    = {Jung Wun Lee},
  title     = {A Multivariate Latent Class Profile Analysis with Latent Group},
  howpublished = {EasyChair Preprint 874},
  year      = {EasyChair, 2019}}