gic.csgl: Fits CSGL, choosing lambda by GIC

In aplantin/CSGL: Compositional Sparse Group Lasso

Description

Fits CSGL for a sequence of lambdas and chooses lambda based on the GIC.

Usage

gic.csgl(Z, y, groups, mu = 1, theta = 0.95, lam.seq = NULL, nlam = 25,
  min.frac = 0.1, thresh = 1e-07, maxit = 10000, std = T, verbose = F)

Arguments

Z	Matrix of log OTU proportions
y	Outcome vector
groups	Vector indicating group membership of each OTU
mu	Augmented Lagrangian parameter
theta	Defines convex combination between L1 and L2 penalties (theta = 1 is L1 only, theta = 0 is L2 only)
lam.seq	Sequence of lambda values to consider
nlam	Number of lambda values to try
min.frac	Minimum value of the penalty parameter, as a fraction of the maximum value
thresh	Threshold for convergence
maxit	Maximum number of iterations
std	Logical flag for variable standardization before fitting the model
verbose	Logical flag for printing updates during model fitting

Value

A list with components

gic.int	Intercept for lambda chosen by GIC
gic.beta	Coefficient vector for lambda chosen by GIC
fullfit	Full model fit (all lambda values)
gic	Vector of GIC values for each lambda

aplantin/CSGL documentation built on May 21, 2019, 12:08 p.m.