.glmnetRandomizedLasso {monaLisa}R Documentation

Randomized Lasso

Description

This function performs randomized lasso using the glmnet package. The function present in the stabs package that runs the lasso version was adapted for the randomized lasso here. Randomized lasso stability selection uses this function repeatedly to select predictors.

Usage

.glmnetRandomizedLasso(
  x,
  y,
  q,
  weakness = 1,
  type = c("conservative", "anticonservative"),
  ...
)

Arguments

x

the predictor matrix. Passed to x of glmnet.lasso from stabs package.

y

the response vector. Passed to y of glmnet.lasso from stabs package.

q

the number of variables that are selected on each subsample. Passed to q of glmnet.lasso from stabs package.

weakness

weakness parameter used in randomized lasso (see details).

type

parameter passed to type of glmnet.lasso from stabs package. It is a character vector specifying how much the PFER should be controlled. If type is "conservative" (default), then the number of selected variables per subsample is <= q. If type is "anticonservative" then the number of selected variables per subsample is >= q.

...

additional parameters for glmnet.

Details

This function is identical to glmnet.lasso from the stabs package. The only addition/modification is the weakness parameter which has been added when calling the glmnet function by setting penalty.factor = 1/runif(ncol(x), weakness, 1), where ncol(x) is the number of predictors.

Value

the regression output which consists of a list of length 2. The list contains the following:

See Also

glmnet.lasso and glmnet


[Package monaLisa version 0.99.5 Index]