stability.MLGL: Stability Selection for Multi-Layer Group-lasso

Description

Stability selection for MLGL

Usage

stability.MLGL(
  X,
  y,
  B = 50,
  fraction = 0.5,
  hc = NULL,
  lambda = NULL,
  weightLevel = NULL,
  weightSizeGroup = NULL,
  loss = c("ls", "logit"),
  intercept = TRUE,
  verbose = FALSE,
  ...
)

Value

a stability.MLGL object containing:

lambda: sequence of lambda.
B: Number of bootstrap samples.
stability: A matrix of size length(lambda)*number of groups containing the probability of selection of each group
var: vector containing the index of covariates
group: vector containing the index of associated groups of covariates
time: computation time

Arguments

X: matrix of size n*p
y: vector of size n. If loss = "logit", elements of y must be in -1,1
B: number of bootstrap sample
fraction: Fraction of data used at each of the B sub-samples
hc: output of hclust function. If not provided, hclust is run with ward.D2 method
lambda: lambda values for group lasso. If not provided, the function generates its own values of lambda
weightLevel: a vector of size p for each level of the hierarchy. A zero indicates that the level will be ignored. If not provided, use 1/(height between 2 successive levels)
weightSizeGroup: a vector
loss: a character string specifying the loss function to use, valid options are: "ls" least squares loss (regression) and "logit" logistic loss (classification)
intercept: should an intercept be included in the model ?
verbose: print some informations
...: Others parameters for gglasso function

Author

Quentin Grimonprez

Details

Hierarchical clustering is performed with all the variables. Then, the partitions from the different levels of the hierarchy are used in the different runs of MLGL for estimating the probability of selection of each group.

References

Meinshausen and Buhlmann (2010). Stability selection. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 72.4, p. 417-473.

Examples

Run this code

# \donttest{
set.seed(42)
# Simulate gaussian data with block-diagonal variance matrix containing 12 blocks of size 5
X <- simuBlockGaussian(50, 12, 5, 0.7)

# Generate a response variable
y <- X[, c(2, 7, 12)] %*% c(2, 2, -2) + rnorm(50, 0, 0.5)

# Apply stability.MLGL method
res <- stability.MLGL(X, y)
# }

Run the code above in your browser using DataLab

Last chance! 50% off unlimited learning