Estimating sparse networks with hubs (Q2196140)

The authors investigate the problem of estimating sparse networks in which there are a few highly connected hub nodes. A new method for estimating networks with hubs is proposed that exploits the ability of inverse covariance estimation methods to include structural information about the underlying network. A practical advantage of the new method is that it leads to an optimization problem that is solved using the efficient graphical Lasso algorithm that is already implemented in the R package glasso. The idea of Lasso algorithm is to estimate the precision matrix by the maximizer of the penalized log-likelihood function. The authors establish large sample properties of the method when the number of parameters diverges with the sample size. It is shown via simulations that the method outperforms competing methods and its performance is illustrated with an application to microbiome data.