Functional Data Analytic Approaches to Identifying Biosignatures Based on Imaging Data

• Activity Number: 340
• Type: Topic Contributed
• Date/Time: Tuesday, August 6, 2013 : 10:30 AM to 12:20 PM
• Sponsor: Mental Health Statistics Section
• Author(s): R. Todd Ogden*+
• Companies: Columbia University
• Keywords: functional data ; biosignature ; wavelets

Abstract

In many biomedical applications it is of interest to use imaging data or other very high dimensional data as predictors in regression models, e.g., to predict a patient's treatment outcome based on brain imaging data obtained at baseline. Obtaining meaningful fits in such problems requires some form of dimension reduction while taking into account the structure of the data, a primary goal of functional data analysis. This talk will discuss various functional data analytic approaches for fitting such a model. In particular, wavelet analysis provides a set of useful tools that can allow regression models to focus on important aspects of the images across a range of scales.