Combinations of covariance selections for graphical modelling.

We explore the possibility of composing the results of a fixed number of Gaussian graphical model selections on some partially overlapping variables. This appears to be an useful approach in all the research areas where a large amount of data from different sources and types of experiments is available. Therefore the focus is in binding together information coming from heterogeneous studies to improve the understanding of a particular phenomenon of interest. The proposed approach relies on numerical results on artificial and real data.