For each class, returns lists of all features belonging to subnetworks. (A subnetwork is defined as a collection of features C for which theta[C,!C]==0, and within which no further subnetworks can be identified. In other words, a block in the block diagonal structure of theta, or a set of features that can be connected through theta's edges.)

1 | ```
subnetworks(theta)
``` |

`theta` |
A list of pXp matrices, each an estimated sparse inverse covariance matrix. (For example, the result of FGL or GGL.) |

A list length K, each element of which is a list of subnetworks in class K. Each subnetwork is represented as a vector of feature names.

Patrick Danaher

Patrick Danaher, Pei Wang and Daniela Witten (2011). The joint graphical lasso for inverse covariance estimation across multiple classes. http://arxiv.org/abs/1111.0324

1 2 3 4 5 6 7 | ```
## load an example dataset with K=two classes, p=200 features, and n=100 samples per class:
data(example.data)
str(example.data)
## run fgl:
fgl.results = JGL(Y=example.data,penalty="fused",lambda1=.25,lambda2=.1)
## identify subnetworks
subnetworks(fgl.results$theta)
``` |

Questions? Problems? Suggestions? Tweet to @rdrrHQ or email at ian@mutexlabs.com.

All documentation is copyright its authors; we didn't write any of that.