Bayesian Hierarchical Model for Large-Scale Covariance Matrix Estimation

Many bioinformatics problems can implicitly depend on estimating large-scale covariance matrix. The traditional approaches tend to give rise to high variance and low accuracy estimation due to “overfitting”, and hence not completely satisfactory. We cast the large-scale covariance matrix estimation problem into the Bayesian hierarchical model framework, and introduce dependency between covariance parameters. We demonstrate the advantages of our approaches over the traditional approaches using simulations and an exemplary omics data analysis. Estimating covariance matrix from high-throughput “omics” data is indispensable for many tasks, notably for finding clusters in the data, whether of the hierarchical or network flavor. The problem remains to be challenging due to the large number of variables p (such as genes or proteins) and the comparatively small number of samples n (such as conditions under which gene expression is measured). The existing approaches that rely on the...

• Type: Journal
• Year: 2007
• Where: JCB
• Authors: Dongxiao Zhu, Alfred O. Hero III
• Keywords: Covariance Matrix, Estimation, JCB 2007, Large-scale Covariance Matrix