External Seminar: Richard Boys (Newcastle)

[Statistics & Probability Seminar]

Revealing subgroup structure in ranked data using a Bayesian WAND

Ranked data arise in many areas of application ranging from the ranking of up-regulated genes for cancer to the ranking of academic statistics journals. Complications can arise when rankers don't report a full ranking of all entities; for example, they might only report their top 10 ranked entities after seeing some or all entities. It can also be useful to know whether rankers are equally informative and whether some entities are effectively judged to be exchangeable. Recent work has looked at how to produce an aggregate (overall) ranking but this is not a very useful summary when there is important subgroup structure. This talk will outline some current work studying this problem using Bayesian non-parametrics, specifically, a Weighted Adapted Nested Dirichlet (WAND) process mixture of Plackett-Luce models. Fortunately an efficient Gibbs sampling scheme is available after introducing some latent variables. The method will be illustrated using real world examples available in the literature.