plsDA_main: plsDA_main partial least squares discriminant analysis

In zdk123/compPLS: Conduct Partial Least Squares regression for compositional data

plsDA_main partial least squares discriminant analysis

Description

The main wrapper for full Partial Least Squares discriminant analysis, performing cross-validation to tune model parameters (here, number of components) and do permutation tests (ie bootstrapping) to get pseudo-pvals estimates for model coefficients

The main wrapper for full sparse Partial Least Squares discriminant analysis, performing cross-validation to tune model parameters (here, number of components) and do permutation tests (ie bootstrapping) to get pseudo-pvals estimates for model coefficients

Usage

plsDA_main(
  x,
  grouping,
  K,
  usePriors = FALSE,
  fold = 5,
  nboots = 999,
  n.core = 4,
  noise = 0,
  ...
)

splsDA_main(
  x,
  grouping,
  eta,
  K,
  usePriors = FALSE,
  fold = 5,
  nboots = 999,
  n.core = 4,
  noise = 0,
  ...
)

Arguments

x	data with samples in rows, features are columns (not necessarily compositional x)
grouping	a numeric vector or factor with sample classes (length should equal nrow(x))
K	numeric vector containing number of components in the PLS model
usePriors	use priors for very biased sample size between groups (ie - put strong penalty on misclassifying small groups)
fold	number of partitions to randomly subsample for cross-validation
nboots	number of bootstraps/permutations for estimating coefficient p-vals
n.core	number of cores for paralellization of bootstraps
noise	for very sparse components, some subsamples may have zero variance. Optionally, add some Gaussian noise to to avoid PLS errors
...	additional arguments passed to plsDA

Value

a plsDA object that contains: the plsda model/object, pvals, the original data, x, and groupings

a plsDA object that contains: the plsda model/object, pvals, the original data, x, and groupings

See Also

plsDA

plsDA

zdk123/compPLS documentation built on April 24, 2022, 2:44 p.m.