Testing and Estimation of Partial Correlation Networks

• Activity Number: 120
• Type: Topic Contributed
• Date/Time: Monday, July 30, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Min Jin Ha*+ and Fred Wright
• Companies: and The University of North Carolina at Chapel Hill
• Keywords: partial correlation ; ridge estimation ; simultaneous hypothesis testing

Abstract

We describe a statistical framework to estimating partial correlation networks, while simultaneously performing valid hypothesis testing for network edges. We perform simple modifications of sample correlation matrices to handle singularities in estimating partial correlations, utilizing a 'best of breed' approach to identify the best among a series of estimation procedures applied to a particular dataset. Challenges in estimation of standard errors and construction of p-values are handled using resampling approaches. Extensive simulation studies demonstrate the performance of various estimation methods under various scenarios, enabling methods with high accuracy to be chosen for each dataset. Our graph construction procedure is applied to yeast cell cycle microarray expression data, and the results validated using known pathways.