Multiscale Adaptive Composite Quantile Regression Models for Neuroimaging Data

• Activity Number: 310
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Graphics
• Author(s): Linglong Kong*+ and Hongtu Zhu
• Companies: The University of North Carolina at Chapel Hill and The University of North Carolina
• Keywords: Kernel ; Multiscal adaptive regression ; Neuroimaging data ; Propagation separation ; composite quantile regression ; Robustness

Abstract

Neuroimaging studies aim to analyze imaging data with complex spatial patterns in a large number of locations (called voxels) on a two-dimensional (2D) surface or in a 3D volume. We propose a multiscale adaptive composite quantile regression model (MACQRM) that has four attractive features: being robustness, being spatial, being hierarchical, and being adaptive. MACQRM utilizes imaging observations from the neighboring voxels of the current voxel and borrows strength from the nearby quantile regressions of the current regression to adaptively calculate parameter estimates and test statistics. Theoretically, we establish consistency and asymptotic normality of the adaptive estimates and the asymptotic distribution of the adaptive test statistics. Our simulation studies and real data analysis confirm that MACQRM significantly outperforms MARM and conventional analyses of imaging data.