A Finite Mixture Multinomial Probit Model for Choice Based Conjoint Analysis: A Simulation Study

In the context of conjoint choice models the use of finite mixture models to estimate and analyze segment-specific consumer preferences is well-established. However, finite mixture models usually do not account for different pairwise similarities of alternatives, which could lead to biased estimates and predictions. In this contribution, we develop a Finite Mixture MNP model that is able to account for such dependencies, and assess its performance in terms of model fit, parameter recovery and forecasting accuracy. As a benchmark for comparison, we use the Finite Mixture IP model which belongs to the same model class but instead assumes independence between alternatives. Our results indicate a significantly better performance of the new Finite Mixture MNP model with respect to (unpenalized) model fit and parameter recovery.