ISSN 1671-3710
CN 11-4766/R
主办:中国科学院心理研究所
出版:科学出版社

Advances in Psychological Science ›› 2017, Vol. 25 ›› Issue (10): 1696-1704.doi: 10.3724/SP.J.1042.2017.01696

• Research Methods • Previous Articles     Next Articles

 Piecewise growth mixture models and its current researches

 WANG Jing1,2; TANG Wenqing5; ZHANG Minqiang1,2,3,4; ZHANG Wenyi6; GUO Kaiyin4   

  1.  (1 School of Psychology, South China Normal University) (2 Center for Studies of Psychological Application, South China Normal University) (3 Guangdong Key Laboratory of Mental Health and Cognitive Science, South China Normal University) (4 Guangdong Psychological Association, Guangzhou 510631, China) (5 School of Education, Guangxi University, Nanning 530004, China) (6 School of Entrepreneurship Education, Guangdong University of Finance & Economics, Guangzhou 511300, China)
  • Received:2016-10-26 Online:2017-10-15 Published:2017-08-13
  • Contact: ZHANG Minqiang, E-mail: zhangmq1117@qq.com E-mail:E-mail: zhangmq1117@qq.com
  • Supported by:
     

Abstract:  Piecewise growth mixture models (PGMM) can be used to analyze multi-phase longitudinal data with unobserved heterogeneity in a population, and are widely applied in fields such as ability growth, social behaviors development and intervention, and clinical psychology. PGMM can be defined within both the structural equation modeling framework and the random coefficient modeling framework. Maximum likelihood via an expectation–maximization algorithm (EM-ML) and Markov Chain Monte Carlo for Bayesian inference (MCMC-BI) are the most commonly used methods for PGMM parameter estimation. The validity of PGMM and their parameter estimation are significantly affected by factors such as sample size, number of time points, and latent class separation. Future studies should focus on comparisons between PGMM and other growth models, and the influences of factors such as data characters and latent class attributes on the performance of parameter estimation methods under the same modeling framework or different modeling frameworks.

Key words: longitudinal data, growth mixture models, piecewise growth mixture models, parameter estimation methods

CLC Number: