Title: Networks: Sparse penalized estimation methods using elastic-net penalty
Authors: Davide Bernardini - University of Trento (Italy) [presenting]
Emanuele Taufer - University of Trento (Italy)
Sandra Paterlini - University of Trento (Italy)
Abstract: In the context of Gaussian Markov networks, three estimators for graphical models with an elastic-net penalty are developed and tested. The goal is to estimate the sparse precision matrix from which to retrieve both the underlying conditional dependence graph and partial correlation graph. The first estimator relies on using conditional penalized regressions to estimate the precision matrix. In contrast, the second approach is based on direct penalization of the precision matrix in the likelihood function. Finally, the third estimator relies on a 2-stages procedure that estimates the edge set firstly and then the precision matrix. Through simulations, we investigate the performances of the proposed methods on a large set of well-known network structures. We show how a 2-steps procedure can improve the estimate both of the sparsity pattern in the graph and the edges' weights when we are interested in reconstructing a partial correlation graph.