Bradley-Terry models for pair-comparison networks: Structure and scalability

In this talk I will discuss various aspects of Bradley-Terry models, focusing especially on some aspects that can make the Bradley-Terry model an interesting "toy" case when considering scalable computation and inference more generally.  Included will be some (very!) fresh simulation experiments to investigate: (a) large-network bias; and (b) model reformulation to improve distributional properties as well as computational tractability in "structured" Bradley-Terry models.  This relates to on-going work with Warwick PhD students David Selby and Ella Kaye, and also with Heather Turner (Warwick) and Ioannis Kosmidis (UCL) --- and so I am especially keen to engage with other participants in this INI programme, to explore competing approaches.