|
Abstract:
|
Graphical models have been important tools to depict conditional dependence structure among high-dimensional random variables and have been widely applied in various applications such as gene networks, brain activities, etc. When variables are functional variables, the estimation of the graph structures is more challenging. This problem becomes even more complicated when substantial heterogeneity exists. In this presentation, we propose a mixture functional graphical model to estimate the structures of networks from heterogeneous functional data, which helps to discover hidden patterns that would be ignored and reduces bias estimation. The benefits of our method are demonstrated through numeric analysis both in simulated data and in real data analysis of brain imaging of autism spectrum disorder.
|
