GVARCLUS Procedure

The GVARCLUS procedure performs variable clustering and graphical modeling in SAS Viya. The procedure divides a set of variables into disjoint clusters and creates tables that contain the edge and vertex information for defining an undirected graph. Variables in different clusters are conditionally independent given their own clusters. The procedure also provides the edge and vertex information for an undirected graphical model; this information expresses the relationships among all the variables. A regularization parameter is used to control the sparsity of connections among variables. Tuning the regularization parameter from low to high increases the number of disconnected components and splits larger clusters into smaller ones. Those divided clusters naturally form a hierarchical structure during this process.

The GVARCLUS procedure helps mine the relationships among variables and remove redundant variables. Removing redundant variables is especially important for high-dimensional data, which are increasingly common in the modern data mining world.