Closed-form random utility models with mixture distributions of random utilities: Exploring finite mixtures of qGEV models

This paper investigates the class of random utility models (RUM) derived under the assumption of random utilities or disutilities following mixture distributions. We introduce a general framework embedding the models of Mattsson et al. (2014) and the q-product GEV model of Chikaraishi and Nakayama, (2016) while extending the investigations of Papola (2016). New closed-form models are obtained, with mixtures of covariance matrices, mathematical forms of utility (additive, multiplicative or in-between), variances of utilities (heteroscedasticity) and marginal distributions. The models are compared in two cross-validation exercises, based on a real dataset of travel mode preferences, outperforming existing heteroscedastic and homoscedastic closed-form models in terms of both in-sample and out-of-sample goodness of fit. The behavioural implications of the models are also discussed.