Archetypal analysis and latent Markov models: A step-wise approach

In this contribution we exploit probabilistic archetypal analysis in a three-step approach involving latent Markov models to analyze discrete latent variables with a discrete-time follow-up scheme. Archetypal analysis provides class assignments that are subsequently used as single indicators in a latent Markov model to estimate the structural part. We apply the proposed strategy to a dataset concerning responses to a statistical anxiety questionnaire administered to university students attending an introductory statistical course.