Integrative Analysis of Incompatible High-Dimensional Data Sets with Different Resolutions

• Activity Number: 673
• Type: Topic Contributed
• Date/Time: Thursday, August 13, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Yuan Jiang*
• Companies: Oregon State University
• Keywords: Genetic Association Studies ; Homogeneity ; Incompatible Datasets ; Integrative Analysis ; Resolution ; Statistical Regularization

Abstract

As advanced platforms produce data in a higher and higher resolution, a large collection of high-dimensional datasets are made available from different studies with possibly various resolutions. These multi-resolution datasets are incompatible in the sense that a predictor in a low-resolution dataset corresponds to a group of predictors in a high-resolution dataset. This incompatible but nested data structure poses new challenges to the existent statistical methods for data integration. This paper proposes a statistical regularization approach that can integratively analyze multiple high-dimensional datasets with different resolutions. This approach not only enables joint estimation of model parameters but also ensures consistent findings from multiple studies. Simulation studies illustrate the advantage of the proposed joint analysis in terms of its consistent findings and its enhanced statistical power compared to separate analyses. Meanwhile, an integrative analysis of multi-resolution genetic datasets shows the applicability of the proposed method to genetic association studies.