Robust discrete choice models with t-distributed kernel errors

Outliers in discrete choice response data may result from misclassification and misreporting of the variable behaviour that is inconsistent with modelling assumptions (e.g. random utility maximisation). In presence outliers, standard models produce biased estimates suffer compromised predictive accuracy. Robust statistical are less sensitive to outliers than non-robust models. This paper analyses two robust alternatives multinomial probit (MNP) model. The robit whose kernel error distributions heavy-tailed t-distributions moderate influence outliers. first model (MNR) model, which a generic degrees freedom parameter controls heavy-tailedness distribution. second generalised (Gen-MNR) more flexible MNR, as it allows for distinct each dimension For both models, we derive Gibbs samplers posterior inference. simulation study, illustrate excellent finite sample properties proposed Bayes estimators show MNR Gen-MNR accurate if contain through lens MNP case study on transport mode behaviour, outperform by substantial margins terms in-sample fit out-of-sample also highlights differences elasticity across