|
Abstract:
|
Understanding brain networks is necessary for developing improved diagnosis and treatment strategies for mental health disorders. In order to better understand how the brain network varies according to phenotypes (behavioral traits, neurological disorders, etc.), novel statistical methods are needed for analyzing network-valued data consisting of a different network for each individual. We develop a Bayesian semiparametric model, which combines low-rank factorizations and Gaussian process priors to allow flexible shifts of the conditional expectation for a network-valued random variable across the values of a predictor, while including subject-specific random effects to improve prediction. The formulation leads to a simple Gibbs sampler and we demonstrate the good performance of our model in prediction and goodness-of-fit assessments. The model is applied to learn changes in the brain network across intelligence scores and we find that intelligence score is positively correlated with average connection probability of potential edges connecting the left and right frontal lobes.
|
