Title: Latent class trees
Authors: Mattis van den Bergh - Tilburg University (Netherlands) [presenting]
Jeroen Vermunt - Tilburg University (Netherlands)
Abstract: Researchers use latent class analysis to derive meaningful clusters from sets of categorical observed variables. However, especially when the number of classes required to obtain a good fit is large, interpretation of the latent classes in the selected model may not be straightforward. To overcome this problem, we propose an alternative way of performing a latent class analysis, which we refer to as latent class tree modelling. For this purpose, we use a recursive partitioning procedure similar to those used in divisive hierarchical cluster analysis; that is, classes are split until the model selection criterion indicates that the fit does no longer improve. The key advantage of the proposed latent class tree approach compared to the standard latent class analysis approach is that it gives a clear insight into how the latent classes are formed and how solutions with different numbers of classes are linked to one another. We also propose measures to adjust the tree in certain conditions. The new approach is illustrated by the analysis of real data and simulation studies.