mcl_step_cv: mcl_step_cv

In aplantin/MCL: Fits multilevel compositional lasso

Description

Cross-validated choice of lam1 and lam2 (over a grid of lambdas)

Usage

mcl_step_cv(Z, W, y, mu = 1, groups, step = 0.1, lam.seq = NULL,
  lam.max = NULL, nlam = c(10, 10), min.frac = 0.01, nfolds = 5,
  maxit = 10000, thresh = 1e-04, std = TRUE, verbose = 1)

Arguments

Z	Otu matrix at group level
W	Otu matrix within groups
y	Outcome vector
mu	Lagrange scaling parameter, default 1
groups	Vector indicating group membership for each feature
step	Step-size, default 0.1
lam.seq	Sequence of lambda values, default NULL
lam.max	Maximum lambda value, default NULL
nlam	2-vector of number of lambdas to test (lam1 and lam2), default (10, 10)
min.frac	Fraction of max.lam that the minimum lambda value should be, default 0.01
nfolds	Number of folds for cross validation, default 5
maxit	Maximum number of iterations, default 10000
thresh	Threshold for convergence, default 0.0001
std	Logical indicating whether data should be standardized, default TRUE
verbose	If 0, prints nothing; if 1, prints which fold you're on; if 2, also prints a progress bar for lambdas. Default 1

Value

List: full fit, lambda sequence, CV-selected lam1, CV-selected lam2, cross-validated prediction error

aplantin/MCL documentation built on May 25, 2019, 2:26 p.m.