Latent Class Cluster Analysis: Selecting the number of clusters.

Area of Science:

• Social Sciences
• Psychology
• Educational Research

Background:

• Latent Class Cluster Analysis (LCCA) is a powerful model-based clustering technique.
• Determining the correct number of clusters in LCCA is complex and often subjective.
• Current reliance on fit indices alone can lead to suboptimal cluster selection.

Purpose of the Study:

• To propose an extended analytical strategy for selecting the number of clusters in LCCA.
• To improve the objectivity and reliability of cluster number selection in LCCA.
• To provide a reproducible method for LCCA implementation.

Main Methods:

• An extended strategy integrating model fit, cluster separation, and partition stability.
• Illustration with simulated data and a real-world dataset (ICILS 2018).
• Provision of an R script for end-to-end LCCA, including data preprocessing.

Main Results:

• The proposed strategy facilitates the selection of more stable and well-separated clusters compared to using only fit indices.
• Demonstrated effectiveness on both simulated and real-world data.
• The strategy aids researchers in identifying generalizable clusters.

Conclusions:

• The extended strategy offers a more robust approach to determining the number of clusters in LCCA.
• This method enhances the quality and interpretability of LCCA findings.
• Researchers can apply this strategy using various clustering software.