Bayesian nonparametric dynamic modeling of psychological traits

This work focuses on investigating the evolution of different traits of psychosis during the COVID-19 pandemic. We develop a Bayesian nonparametric mixture model for multivariate categorical data, which characterizes the population’ psychosis via a set of latent psychological profiles. Leveraging a time- and covariate- dependent stick-breaking construction for the mixture weights, the proposed specification characterizes the dynamic evolution of such latent traits across the pandemic, measuring the effect of subject-specific demographic information such as sex and age of the individuals.