Accounting for model uncertainty in individualized designs for discrete choice experiments

We consider the issue of finding optimal designs for Discrete Choice Experiments (DCE). Currently the optimal design for a DCE is predicated on a specific mixed logit model which represents the probability of choosing a specific alternative in a choice set through a linear predictor combining several factors. Using a unique model represents a major limitation which we overcome by allowing for a collection of different models characterized by distinct linear predictors. In this new setting, model choice becomes a major concern. By maximizing the utility of selecting a choice set - computed as the mutual information between the model indicator and the predicted observation at that set - over alternative choice sets, model discrimination is enhanced. We implement our methodology using a sequential Monte Carlo algorithm suitably tailored to deal with model uncertainty.