B0724
Title: BART for network-linked data
Authors: Sameer Deshpande - University of Wisconsin--Madison (United States) [presenting]
Abstract: Regression with network-linked data is considered in which (1) covariate-response pairs are observed at the vertices of a given network but (2) the regression relationship might be different vertex-to-vertex. It describes how to use the popular Bayesian additive regression trees (BART) model for this problem in a way that does not require pre-specifying the functional form of the regression function or how the regression function varies across the network. Key to the proposal are several stochastic processes that randomly partition a network into two, possibly connected, components.
