Double clustering with a matrix-variate regression model: finding groups of athletes and disciplines in decathlon’s data

Decathlon is a two-day track and field competition in which athletes compete in 10 different disciplines placed in an invariant order. In this work we propose a co-clustering method using a generalization of the Bayesian matrix-variate regression model, in which both athletes and disciplines are grouped according to the observed data. Specifically, we describe the model and briefly outline the inferential strategy, based on a Gibbs sampling algorithm. The model provides powerful insights that can be used by coaches and athletes to improve their performances