A Stepwise Approach for High-Dimensional Gaussian Graphical Models

We present a stepwise approach to estimate high dimensional Gaussian graphicalmodels. exploit the relation between partial correlation coefficientsand distribution of prediction errors, and parametrize model in termsof Pearson coefficients errors nodes’best linear predictors. propose novel algorithm for detecting pairsof conditionally dependent variables. compare proposed withexisting methods including graphical lasso (Glasso), constrained `l1-minimization(CLIME) equivalent (EPC), via simulation studies andreal life applications. In our study we consider several settingsand report results using different performance measures that look at desirablefeatures recovered graph.