grouplassoLambdas: Compute the lambdas (adaptive group lasso tuning parameters)...

In wrbrooks/lagr: Local adaptive grouped regularization

Description

Finds lambda.max, the smallest lambda that shrinks all the penalized groups to zero. The rest of the lambdas form a log-linear sequence from lambda.max to lambda.min = min.frac*lambda.max.

Usage

grouplassoLambdas(data, index, family, weights, adaweights, min.frac = 0.05,
  nlam = 20, type = "linear")

Arguments

data	list consisting of x, a matrix of grouped lcoal covariates, and y, the vector of observed responses
index	vector indicating group membership for each column of x
weights	vector of observations weights
adaweights	vector of adaptive group penalty multipliers (computed from the unpenalized coefficients)
min.frac	ratio of lambda.max to lambda.min
nlam	how many lambdas to compute along the sequence from lambda.max to lambda.min
type	string indicating the kind of regression model, defaults to "linear"

Value

vector of adaptive group lasso tuning parameters at which to compute the local LAGR coefficients

wrbrooks/lagr documentation built on May 4, 2019, 11:59 a.m.