This thesis explores extensions of Random Utility Models (RUMs), providing more flexible models and adopting a computational perspective. This includes building new models and understanding their properties such as identifiability and the log concavity of their likelihood functions as well as the development of estimation algorithms.
