Skip to contents

Implementation of the CCLasso approach (Fang et al., 2015), which is published on GitHub (Fang, 2016). The function is extended by a progress message.

Usage

cclasso(
  x,
  counts = F,
  pseudo = 0.5,
  sig = NULL,
  lams = 10^(seq(0, -8, by = -0.01)),
  K = 3,
  kmax = 5000,
  verbose = TRUE
)

Arguments

x

numeric matrix (nxp) with samples in rows and OTUs/taxa in columns.

counts

logical indicating whether x constains counts or fractions. Defaults to FALSE meaning that x contains fractions so that rows sum up to 1.

pseudo

numeric value giving a pseudo count, which is added to all counts if counts = TRUE. Default is 0.5.

sig

numeric matrix giving an initial covariance matrix. If NULL (default), diag(rep(1, p)) is used.

lams

numeric vector specifying the tuning parameter sequences. Default is 10^(seq(0, -8, by = -0.01)).

K

numeric value (integer) giving the folds of crossvalidation. Defaults to 3.

kmax

numeric value (integer) specifying the maximum iteration for augmented lagrangian method. Default is 5000.

verbose

logical indicating whether a progress indicator is shown (TRUE by default).

Value

A list containing the following elements:

cov.wCovariance estimation
cor.wCorrelation estimation
lamFinal tuning parameter

References

fang2015cclassoNetCoMi

fang2016cclassoGithubNetCoMi

Author

Fang Huaying, Peking University (R code)
Stefanie Peschel (documentation)