Latent class procedures: Recent development and applications.
Chapter
Overview
Research
Identity
Additional Document Info
View All
Overview
abstract
Statistical methods commonly used in organizational research usually either take a variable-centered approach or a person-centered approach. Statistical methods taking a person-centered approach are used in studies when organizational researchers are interested in classifying individuals into subpopulations that differ from each other in patterns of variables. They include cluster analysis and latent class analysis. Recent development in quantitative methods-has extended the latent class analysis to integrate variable-centered and person-centered analytical approaches. Specifically, the newly developed latent class procedures (e.g., mixed-measurement item response models, growth mixture modeling, and latent mixture Markov modeling) classify individuals into subpopulations, conditional not only on their similarities in patterns of variables, but also on various types of interrelatedness among variables (e.g., item response patterns and longitudinal quantitative and qualitative changes). Latent class procedures have the following features. First, latent class procedures are based on the assumption that unobserved subpopulations exist within the target population. Second, latent class procedures are model-based methods that are different from methods that cluster data with arbitrary data-driven criteria. Third, latent class procedures can estimate the posterior probabilities of individuals belonging to each latent class. Finally, latent class procedures can be used to test the mechanisms underlying the effects from the observed population heterogeneity (e.g., gender, race, or culture). In the following sections of this chapter, we introduce three recently developed latent class procedures in detail. They are mixed-measurement item response models, growth mixture modeling, and mixture latent Markov modeling. For each modeling technique, we first discuss which research questions can be addressed by the technique and then introduce model specification, model estimation, and model selection procedures. (PsycInfo Database Record (c) 2022 APA, all rights reserved)
