Markovian-Based Clustering of Internet Addiction Trajectories

Abstract

The main aim of this paper is to describe the use of the Markovian-based Hidden Mixture Transition Distribution (HMTD) model for the clustering of longitudinal sequences of continuous data. We especially discuss the use of covariates to improve the clustering process. The HMTD is compared to the well-known Growth Mixture Model (GMM) that is considered here as a gold standard. Both models are applied to a sample of n = 185 adolescents, who are repeatedly evaluated for Internet overuse using the Internet Addiction Test (IAT). The best solution provided by the HMTD model has four groups and it uses five covariates. This solution is related to the subjects’ level of emotional well-being, body mass index, gender, and education track, but shows no relation with age. Compared to a GMM clustering, the HMTD solution provides highly interpretable results with quite equilibrate cluster size, while GMM tends to identify very small clusters allowing for less generalization.