Bayesian Functional Single Index-Interaction Model for Cognition Prediction

• Activity Number: 34 - Bayesian Functional and Data Models
• Type: Contributed
• Date/Time: Sunday, July 30, 2017 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Kumaresh Dhara* and Debdeep Pati and Debajyoti Sinha
• Companies: Florida State University and Florida State University and Florida State University
• Keywords: Single Index Model ; Markov Chain Monte Carlo ; Bayesian Statistics

Abstract

Cognition prediction from human connectome data is an important area of research in neuroscience. In this article, we develop a novel approach to model cognitive ability as a function of brain connectivity networks and other non-imaging predictors. The model is based on wavelet transform of the connectivity networks which affects the response via a single -index interaction between the wavelet transform and the predictors. We adopt a Bayesian approach with sparsity favoring prior on the wavelet coefficients with suitable identifiability restrictions. Encouraging simulation results demonstrate the efficacy of the approach. The methods are illustrated on predicting cognition score based on structural MRI images of 850 patients enrolled in the human connectome project.