Two Sample Inference on Populations of Graphical Models: Applications to Multi-Subject Functional Brain Connectivity

• Activity Number: 25
• Type: Topic Contributed
• Date/Time: Sunday, August 3, 2014 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Genevera Allen*+ and Manjari I. Narayan and Steffie Tomson
• Companies: Rice University/Baylor College of Medicine and Rice University and University of California, Los Angeles
• Keywords: Functional Connectivity ; Markov Networks ; Gaussian Graphical Models ; Large-Scale Inference ; Post-Selection Inference ; Neuroimaging

Abstract

Gaussian Graphical Models (GGM) are popularly used in neuroimaging studies based on fMRI, EEG or MEG to estimate functional connectivity, or relationships between remote brain regions. In multi-subject studies, scientists seek to identify the functional brain connections that are different between two groups of subjects, i.e. connections present in a diseased group but absent in controls or vice versa. This amounts to conducting two-sample large scale inference over network edges post graph selection. Current approaches include estimating a network for each subject, and then assuming these networks are fixed, conducting inference for each edge. These approaches, however, fail to account for the variability associated with estimating each subject's graph, thus resulting in high numbers of false positives and low statistical power. By using Resampling and Random penalization to estimate the post graph selection variability, and the proper Random Effects test statistics, we introduce a new procedure, termed R3, that solves these problems. Through simulation studies and a multi-subject fMRI study on autism, we show that R3 offers substantial improvements over current methods.