Online Program Home
  My Program

All Times EDT

Abstract Details

Activity Number: 282 - Complex Functional Data Analysis with Biomedical Applications
Type: Invited
Date/Time: Wednesday, August 5, 2020 : 10:00 AM to 11:50 AM
Sponsor: Section on Nonparametric Statistics
Abstract #309439
Title: Correcting Batch Effects in the Covariance Structures of Spatially-Dependent Multivariate Object
Author(s): Andrew Chen and Russell Shinohara and Haochang Shou*
Companies: University of Pennsylvania and University of Pennsylvania and University of Pennsylvania
Keywords: harmonization; neuroimaging; variance decomposition

Robustly integrating and harmonizing data from multi-site studies has become increasingly critical in order to enhance the power of extracting meaningful markers from complex multivariate data. In medical imaging, site differences due to variations in scanning protocols and preprocessing steps are known to exist and have large impact towards the analytic results. Recently, batch-effect corrections methods such as RAVEL and ComBat have been successfully adopted from genomics to neuroimaging data. However, the existing methods have mostly focused on correcting the mean shifts and scale differences for individual dimension across sites. The remain of site differences in covariance structures could pose a concern in multivariate testing as well as analyses for functional connectivity. We will present simulations based on structural and functional imaging measures to demonstrate the site differences and its impact on statistical inferences. Critically, we propose a new approach that extend the ComBat methodology combined with variance decomposition to remove the spatially-dependent site deviations in the covariance structures.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2020 program