An application of graphical models to fMRI data using the lasso penalty

In this thesis, we study the graphical lasso method and apply it to functional magnetic resonance imaging (fMRI) data. The graphical lasso method enables one to construct undirected sparse graphs between variables of interest. The fMRI data concerns subjects brain activities while they engage in saccadic eye movement tasks. The datasets are collected before and after they practice certain tasks. Using the graphical lasso procedure, we create undirected graphs that display the connections between the different regions of interest (ROI) in the brain. By controlling the regularization parameter in this lasso procedure, we identify which ROIs are more strongly connected than the others. We compare these undirected graphs before and after the practice and also across different practice groups.