Two-Sample Testing on Ranked Preference Data and the Role of Modeling Assumptions