huge (version 0.7)

huge.stars: StARS: Stability Approach to Regularization Selection

Description

Implements the Stability Approah to Regularization Selection (StARS).

Usage

huge.stars(R.path, stars.thresh = 0.1, verbose = TRUE)

Arguments

R.path

A list of multiple solution paths. Each solution path is a list of adjacency matrix of estimated graphs using a subsample.

stars.thresh

The varability threshold in stability selection. The default value is 0.1. An alternative value is 0.05.

verbose

If verbose = FALSE, tracing information printing is disabled. The default value is TRUE.

Value

• The index of the optimal regularization parameter is returned.

Details

The StARS selects the optimal regularization parameter using the variability of solution paths by subsampling. It can be applied to multiple solution paths generated by subsampling. An alternative threshold 0.05 is chosen under the assumption that the model is correctly specified. In applications, the model is usually misspecified, 0.1 is a better choise for a safer selection.

References

Tuo Zhao and Han Liu. HUGE: A Package for High-dimensional Undirected Graph Estimation. Technical Report, Carnegie Mellon University, 2010 Han Liu, Kathryn Roeder and Larry Wasserman. Stability Approach to Regularization Selection (StARS) for High Dimensional Graphical Models. Advances in Neural Information Processing Systems, 2010.

See Also

huge.select,huge and huge-package

Examples

#generate data
n = 100
L = huge.generator(n=100, graph = "hub")

#subset indices
ind.group = c(1:30)

#preselect the neighborhood
ind.mat = huge.scr(L$data)$ind.mat
lambda = exp(seq(log(1),log(0.1),length=10))

#subsampling and subgraph estimation
R.path = list()
for(k in 1:6){
    ind.sample = sample(c(1:n), floor(n*0.8), replace=FALSE)
    R.path[[k]] = huge.subgraph(L$data[ind.sample,], ind.group = ind.group, 
    ind.mat = ind.mat, lambda = lambda)$path 
}

#stability selection
D = huge.stars(R.path)