Fig. 4From: Learning mixed graphical models with separate sparsity parameters and stability-based model selectionLearned sub-network of gene expression and clinical features connected to lung disease diagnosis, lung tests and cigarette smoking. Nodes are colored by data type, blue for gene expression, red for clinical variables. Edges were filtered by weight with a threshold of .05. Node size is proportional to the diagonal of the β matrix for continuous variables and ϕ yyF for each categorical variable, yBack to article page