Title: Model-selection for ranking data models
Authors: Lucy Small - University College Dublin (Ireland)
Thomas Brendan Murphy - University College Dublin (Ireland) [presenting]
Abstract: Mallows model is commonly used to model rank data, most often using a form based on Kendall's distance. Generalised Mallows Models (GMM) were developed more recently which expand on the original model. We fit mixtures of these GMMs to rank data, using a Metropolis within Gibbs algorithm. We consider the issue of selecting the most appropriate GMM in practice using a number of different approaches. Applications to simulated data, voting data and wine tasting data are considered.