lol (version 1.20.0)

lasso.simultaneous: Simultaneous lasso

Description

The function performs lasso with multiple random sample splits, selecting coefficients that are simultaneously non-zero in both subsets of samples.

Usage

lasso.simultaneous(y, x=NULL, model='linear', nSubsampling=200, alpha=.5, lambda1=NULL, track=FALSE, ...)

Arguments

y

A vector of gene expression of a probe, or a list object if x is NULL. In the latter case y should a list of two components y and x, y is a vector of expression and x is a matrix containing copy number variables

x

Either a matrix containing CN variables or NULL

model

which model to use, one of "cox", "logistic", "linear", or "poisson". Default to 'linear'

nSubsampling

The number of random permutations, both on sample spliting and on variable scaling, default to 200.

alpha

weakness parameter: control the shrinkage of regulators. The lower alpha is, the bigger the vanishing effect on small coefficients.

lambda1

minimum lambda, default to NULL

track

logical value, whether to track the progress

...

Other parameters to be passed to the penalized function

Value

A list object of class 'lol', consisting of:

beta

Coefficient vector

n

Number of actual subsampling, should be equal or smaller than nSubsampling in case of failing.

mat

result matrix of the subsampling

Details

In each run the function splits samples randomly to two equal sets, run lasso on both sets, then select those coefficients that are simultaneously non-zero across two sets. Finally the results across many runs are summarized as the frequency of selected predictors - the higher the frequency the more confidence that the corresponding predictors are significant.

References

N. Meinshausen and P. Buehlmann (2010), Stability Selection (with discussion), Journal of the Royal Statistical Society, Series B, 72, 417-473.

See Also

lasso

Examples

data(chin07)
data <- list(y=chin07$ge[1,], x=t(chin07$cn))
res <- lasso.simultaneous(data, nSubsampling=50)
res