Online Program Home
  My Program

Abstract Details

Activity Number: 183 - SPEED: Bayesian Methods Student Awards
Type: Contributed
Date/Time: Monday, July 31, 2017 : 10:30 AM to 11:15 AM
Sponsor: Mental Health Statistics Section
Abstract #325149
Title: Granger Mediation Analysis of Functional Magnetic Resonance Imaging Time Series
Author(s): Yi Zhao* and Xi Luo
Companies: Brown University and Brown University
Keywords: SEM ; Confounding ; Sptio-temporal dependence ; feedback effect
Abstract:

Making inference about brain effective connectivity is of great interest in task fMRI experiments. Here, we are interested in clarifying the causal mechanisms of an external stimulus on the outcome brain region, by considering another brain region as an intermediate variable. Causal mediation analysis under structural equation modeling is considered. Attaining causal interpretations requires both the "no unmeasured confounding" and the "no interference" assumptions, which generally do not hold in fMRI datasets. To address the existence of unmeasured confounding, a correlation between model errors is introduced; and to characterize the spatiotemporal dependency, the principle of Granger causality is implemented. In this paper, we propose a Granger Mediation Analysis framework that provides inference about both spatial and temporal causality between brain regions for multilevel fMRI time series. Simulation studies show that our method reduces the estimation bias compared to existing approaches. Applying on a real fMRI dataset, our approach not only estimates the causal effects of brain pathway, but effectively captures the feedback effect of the outcome region on the mediator region.


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

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association