cv.lasso {EAinference}

Compute K-fold cross-validated mean squared error for lasso

Description

Computes K-fold cross-validated mean squared error to propose a lambda value for lasso, group lasso, scaled lasso or scaled group lasso.

Usage

cv.lasso(X, Y, group = 1:ncol(X), weights = rep(1, max(group)), type,
  K = 10L, minlbd, maxlbd, num.lbdseq = 100L, parallel = FALSE,
  ncores = 2L, plot.it = FALSE, verbose = FALSE)

Arguments

X

predictor matrix.

Y

response vector.

group

p x 1 vector of consecutive integers describing the group structure. The number of groups should be the same as max(group). Default is group = 1:p , where p is number of covariates. See examples for a guideline.

weights

weight vector with length equal to the number of groups. Default is rep(1, max(group)).

type

type of penalty. Must be specified to be one of the following: "lasso", "grlasso", "slasso" or "sgrlasso", which correspond to lasso, group lasso, scaled lasso or scaled group lasso.

K

integer. Number of folds

minlbd

numeric. Minimum value of the lambda sequence.

maxlbd

numeric. Maximum value of the lambda sequence.

num.lbdseq

integer. Length of the lambda sequence.

parallel

logical. If parallel = TRUE, uses parallelization. Default is parallel = FALSE.

ncores

integer. The number of cores to use for parallelization.

plot.it

logical. If true, plots the squared error curve.

verbose

logical.

Value

lbd.min

a value of lambda which gives a minimum squared error.

lbd.1se

a largest lambda within 1 standard error from lbd.min.

lbd.seq

lambda sequence.

cv

mean squared error at each lambda value.

cvsd

the standard deviation of cv.

Examples

set.seed(123)
n <- 30
p <- 50
group <- rep(1:(p/10),each=10)
weights <- rep(1, max(group))
X <- matrix(rnorm(n*p),n)
truebeta <- c(rep(1,5),rep(0,p-5))
Y <- X%*%truebeta + rnorm(n)

# To accelerate the computational time, we set K=2 and num.lbdseq=2.
# However, in practice, Allowing K=10 and num.lbdseq > 100 is recommended.
cv.lasso(X = X, Y = Y, group = group, weights = weights, K = 2,
type = "grlasso", num.lbdseq = 2, plot.it = FALSE)
cv.lasso(X = X, Y = Y, group = group, weights = weights, K = 2,
type = "sgrlasso", num.lbdseq = 2, plot.it = FALSE)