Bayesian Clustering of Brain Regions with Three-Dimensional Spatial Covariates

Network data often come with node-level covariates that can be leveraged, for example, in node clustering. In particular, in brain networks, each node – which usually represents a brain region – has three-dimensional anatomical coordinates that can be employed as covariates. However, it is unclear whether all three dimensions are informative for clustering. In this paper, we extend our previous work on Bayesian nonparametric clustering of network nodes by allowing each covariate dimension to be individually included or excluded. The proposed methodology is illustrated on publicly available brain data.