This work proposes novel network analysis techniques for multivariate time series. We define the network of a multivariate time series as a graph where vertices denote the components of the process and edges denote non-zero long run partial correlations. We then introduce a two step lasso...