Time-Varying Networks Estimation and Dynamic Model Selection

• Activity Number: 292
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Xinxin Shu*+ and Annie Qu
• Companies: University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign
• Keywords: Dynamic networks ; local quadratic approximation ; model selection consistency ; proximal gradient method ; varying-coefficient model ; B-spline

Abstract

In many applications, it is important to identify and predict the dynamic changes of associations among network data over time. We propose a varying-coefficient model to incorporate time-varying network data, and impose a piece-wise penalty function to capture local features of the network associations. The advantages of the proposed approach are that it is nonparametric and therefore flexible in modeling dynamic changes of association for network data problems, and capable of identifying the time regions when dynamic changes of associations occur. To achieve local sparsity of network estimation, we implement a group penalization strategy involving overlapping parameters among different groups. But this imposes great challenges in the optimization process for handling large-dimensional network data observed at many time points. We develop a fast algorithm based on the smoothing proximal gradient method, which is computationally efficient and accurate. We illustrate the proposed method through simulation studies and children's attention deficit hyperactivity disorder fMRI data, and show the proposed method and algorithm efficiently recover the dynamic network changes over time.