Estimation of Symmetry-Constrained Gaussian Graphical Models: Application to Clustered Dense Networks

• Activity Number: 148
• Type: Invited
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: SSC
• Author(s): Xin Gao*+ and Helene Massam
• Companies: York University and York University
• Keywords: model selection ; penalized likellihood ; composite likelihood ; Gaussian graphical model ; network analysis ; dense network

Abstract

In this article, we discuss composite likelihood estimation of Gaussian graphical models. When there are symmetry constraints on the concentration matrix or partial correlation matrix, the likelihood estimation can be computational intensive. The composite likelihood offers an alternative formulation of the objective function and the resulting estimation is computationally more convenient. The penalized composite likelihood estimates for edge and vertex class parameters satisfy both symmetry and sparsity constraints and possess ORACLE property. The proposed method can be applied to analyze high-dimensional dense network with large number of edges but sparse edge classes. The empirical performance is demonstrated through simulation studies and a network analysis of a gene expression dataset.