Meeting Banner
Abstract #1718

Granger Causality Via Vector Auto-Regression Tuned for FMRI Data Analysis

Gang Chen1, J. Paul Hamilton2, Moriah E. Thomason2, Ian H. Gotlib2, Ziad S. Saad1, Robert W. Cox1

1Scientific and Statistical Computing Core, NIMH, National Institutes of Health, Bethesda, MD, USA; 2Mood and Anxiety Disorders Laboratory, Department of Psychology, Stanford University, Stanford, CA, USA

We present a platform-independent modeling tool that performs multivariate Granger causality analysis particularly tuned for FMRI data. With an extended vector auto-regressive modeling strategy that accounts for confounds such as baseline drift, head motion parameters, tasks of no interest, physiological measurements, time breaks and signal irregularities, our program provides various model fine-tuning tools including order selections and various diagnosis tests, and identifies the causality at each lag among the pre-selected regions. We also propose a valid group analysis per lag based on signed path coefficients to reveal the network at the group level. The software is in open-source R and available for download.