ldstatsHD (version 1.0.1)

vulLambdaSelection: Vulnerability regularization parameter selection

Description

vulLambdaSelection is a function designed to select the regularization parameter in graphical models. It selects the graph with largest average nodes vulnerability.

Usage

vulLambdaSelection(obj, loo = FALSE, subOut =  10, nite = 50)

Arguments

obj

an object of class huge, camel.tiger or wfgl.

loo

if TRUE an exhaustive leave-one-out procedure is done, otherwise it is used a subsampling approach with nite iterations and leaving out subOut variables.

subOut

number of variables left out in each iteration (only used if loo = FALSE).

nite

number of iterations (only used if loo = FALSE).

Value

An object of class lambdaSelection containing the following components:

opt.lambda

optimal lambda.

crit.coef

coefficients for each lambda given the criterion VUL.

criterion

with value "VUL".

Details

Vulnerability algorithm finds lambda by minimizing the risk function $$ R_{VUL}(\lambda) = - \sum_{i=1}^p \frac{E^\lambda-E_i^\lambda}{E^\lambda} $$ where \(E^\lambda\) is the global efficiency of the original network and \(E^\lambda_i\) is the global efficiency of the network once eliminating the node \(i\). Global efficiency is defined by the harmonic mean of the geodesic distance (see graphDist).

Vulnerability gives \(\lambda\) selection that contains the most vulnerable graph, meaning that the removal of a node in the network in average would affect the most the estimated graph.

References

Costa, L. and F. Rodrigues (2007). Characterization of complex networks: A survey of measurements. Advances in Physics 56(1), 167-242.

See Also

lambdaSelection for other lambda selection approaches.

Examples

# NOT RUN {
# example to use vul function
EX1         <- pcorSimulator(nobs = 50, nclusters = 2, nnodesxcluster = c(40,30), 
                             pattern="powerLaw")
y           <- EX1$y
Lambda.SEQ  <- seq(.35, 0.70, length.out = 10)
out3        <- huge(y, method = "mb", lambda = Lambda.SEQ)
## not run
#VUL.COEF    <- vulLambdaSelection(out3)
#print(VUL.COEF)

 
 
# }