Simultaneous estimation of normal means with side information

The integrative analysis of multiple datasets is an important strategy in data analysis. It increasingly popular genomics, which enjoys a wealth publicly available that can be compared, contrasted, and combined order to extract novel scientific insights. This paper studies stylized example integration for classical statistical problem: leveraging side information estimate vector normal means. task formulated as compound decision problem, oracle rule derived, data-driven this based on minimizing unbiased its risk proposed. shown asymptotically achieve the minimum possible among all separable rules, it outperform existing methods numerical properties. proposed procedure leads naturally high-dimensional classification procedure, illustrated by combining from two independent gene expression profiling studies.