Structure Learning of Multivariate Autoregressive Models based on Marginal Models

Sung-Ho Kim and Yongtae Kim


Multivariate time series, subset of variables, hypothesis testing


Consider a problem of building a vector autoregressive model based on a data set of a limited size which is often the case when analyzing fMRI (functional magnetic resonance imaging) data. We propose a method of structure learning by using the structural information which lies in marginal models. A simulation experiment using a 10-dimensional vector-autoregressive model of order 3 strongly supports the usefulness of the proposed marginal model approach.

Important Links:

Go Back