Think global, act local when estimating a sparse precision matrix

Simulation results suggest that LGM may outperform GLASSO when model sparsity is especially important and when variables in the dataset belong to a number of closely related (if unknown) groups.