GVARCLUS Procedure

supports multiple INPUT statement for both continuous and categorical inputs

provides a FREQ statement for grouped analysis

provides the OUTTREE option to output the hierarchical clustering results in a tree structure

provides the OUTEDGE and OUTVERT options to output the edges and vertices, respectively, for defining an undirected graph

provides the OUTCP option to output a covariance matrix

performs graphical lasso based on Friedman, Hastie, and Tibshirani (2008), which estimates the inverse covariance (IC) matrix at a specified regularization parameter. The inverse covariance matrix interprets the partial correlation between variables given other variables

performs sequential steps of estimating the inverse covariance matrix by using regularization parameters from large to small, which impose different sparsity constraints on the estimation. The sequential steps eventually produce a set of nested clusters that are organized as a hierarchical tree

performs covariance thresholding based on Friedman, Hastie, and Tibshirani (2008) as preprocessing, which thresholds the entries of the sample covariance matrix at the regularization parameter and decomposes the matrix into connected components

enables you to run on a cluster of machines that distribute the data and the computations

enables you to run in single-machine mode

exploits all the available cores and concurrent threads. For information about how PROC GVARCLUS uses threads, see the section Multithreading in Chapter 2, Shared Concepts.