Visualizing Latent Class Models with Analysis-of-distance BIPLOTS. Issue 1 (August 2017)

• Record Type: Journal Article
• Authors: Bakk, Zsuzsa; le Roux, Niel J.
• Is Part Of: Sociological methodology. Volume 47:Issue 1(2017)
• Page Start: 345
• Page End: 378
• Publication Date: 2017-08
• Subjects: biplots -- latent class (LC) models -- multidimensional visualization; Sociology -- Methodology -- Periodicals; Sociology; 301.01
• DOI: 10.1177/0081175017717048
• Languages: English
• ISSNs: 0081-1750

Abstract

The authors propose using categorical analysis-of-distance biplots to visualize the posterior classifications arising from a latent class (LC) model. Using this multivariate plot, it is possible to visualize in two (or three) dimensions the profile of multiple LCs, specifically both the within- and between-class variation, and the overlap or separation of the classes together with the class weights. Furthermore, visualization of the relative density of each of the data patterns associated with a class is possible. The authors illustrate this approach with real data examples of LC models with three and more classes.