zoomJGL {zoomJGL}R Documentation

Estimate zoom-in/out directed and undirected graphs at different coarsness scales

Description

Estimate jointly multiple graphs that resemble each other.

Usage

zoomJGL(Y,lambda1,lambda2,list.controls)

Arguments

Y

List of data matrices where each element of the list is a matrix of dimension Txp, with p increasing across scales.

lambda1

Regularization parameter that dictates the sparsity of the undirected graphs

lambda2

Regularization parameter that dictates the sparsity of the directed graphs

list.controls

List of control parameters specifying

model: The autoregressive model; default AR1

penalize.diagonal: Indicator if the diagonal should be penalized or not; default FALSE

kstar: The starting scale; default 1 (ie. performs a zoom-in)

maxiter: Number of maximal iterations ; default 200

tol: Tolerance threshold for assesing convergence; default 1e-02

rho: Parameter used for the ADMM optimizer; default 1

rho.increments: Parameter used for the ADMM optimizer; default 1

alpha: Parameter used for the ADMM optimizer; default 0.7

Value

A list containing the objects:

Theta

List of parameter matrices of size pxp at each scale. The parameters correspond to undirected edges in the graphs.

Gamma

List of parameter matrices of size pxp at each scale. The parameters correspond to directed edges in the graphs.

Undirected

List of adjacency matrices of size pxp at each scale corresponding to the undirected graphs.

Directed

List of adjacency matrices of size pxp at each scale corresponding to the undirected graphs.

Constraint1

List of matrices that map lambda1 to enforce sparsity and a similar structure across scales for the undirected graphs.

Constraint2

List of matrices that map lambda1 to enforce sparsity and a similar structure across scales for the undirected graphs.

Convergence

Convergence performance at each iteration of the algorithm. The values for DiffObjective, DiffZTheta, DiffGamma which capture the difference between two consecutive iterations for the objective function, the matrices Z-Theta and Gamma should get closer to 0 across iterations.

IC

Information criteria scores: AIC, BIC, eBIC(.1), eBIC(.5), eBIC(.9)

Examples

data(fmri)
obj=zoomJGL(Y,.4,.2)

[Package zoomJGL version 1.0 Index]