Events and Meetings of Italian Statistical Society, Advances in Latent Variables - Methods, Models and Applications

Font Size: 
Simplifying complex latent class modeling using bias corrected three-step approaches
Jeroen Vermunt

Last modified: 2013-06-14

Abstract


Researchers using latent class analysis and other types of mixture models typically proceed using a three-step approach that underestimates the association between class membership and external variables. Some correction methods were proposed in the literature. In this contribution, I will first explain the rather simple maximum likelihood (ML) based correction method. Subsequently, I will show how this three-step approach can be used for the stepwise estimation of complex latent class models, such as latent class models with multiple latent variables, latent Markov models, and multilevel latent class models.

Full Text: PDF