PUB - Publikationen an der Universität Bielefeld

Probit models are widely used to analyze discrete choice behavior in fields such as transportation, marketing, and psychology. We propose a latent class model extension that allows for the classification of decider preferences. The model is estimated in a Bayesian framework, and the number of classes is determined by a Dirichlet process. In a simulation study, we verified that the dependence of the concentration prior diminishes as the number of deciders increases, resulting in stable inference. We further applied the proposed methodology in the context of chess, where chess players are classified according to three types of risk-taking propensity.