cv.plsRmulti: Experimental k-fold cross-validation for multivariate-response PLS2

Description

cv.plsRmulti() performs repeated k-fold cross-validation for the experimental complete-case linear plsRmulti workflow.

Usage

cv.plsRmulti(object, ...)
# S3 method for default
cv.plsRmultiModel(
  object,
  dataX,
  nt = 2,
  limQ2set = 0.0975,
  modele = "pls",
  family = NULL,
  K = 5,
  NK = 1,
  grouplist = NULL,
  random = TRUE,
  scaleX = TRUE,
  scaleY = NULL,
  keepcoeffs = FALSE,
  keepfolds = FALSE,
  keepdataY = TRUE,
  keepMclassed = FALSE,
  EstimXNA = FALSE,
  pvals.expli = FALSE,
  alpha.pvals.expli = 0.05,
  MClassed = FALSE,
  tol_Xi = 10^(-12),
  weights,
  sparse = FALSE,
  sparseStop = FALSE,
  naive = FALSE,
  verbose = TRUE,
  ...
)
# S3 method for formula
cv.plsRmultiModel(
  object,
  data = NULL,
  nt = 2,
  limQ2set = 0.0975,
  modele = "pls",
  family = NULL,
  K = 5,
  NK = 1,
  grouplist = NULL,
  random = TRUE,
  scaleX = TRUE,
  scaleY = NULL,
  keepcoeffs = FALSE,
  keepfolds = FALSE,
  keepdataY = TRUE,
  keepMclassed = FALSE,
  EstimXNA = FALSE,
  pvals.expli = FALSE,
  alpha.pvals.expli = 0.05,
  MClassed = FALSE,
  tol_Xi = 10^(-12),
  weights = NULL,
  subset = NULL,
  contrasts = NULL,
  sparse = FALSE,
  sparseStop = FALSE,
  naive = FALSE,
  verbose = TRUE,
  ...
)

Value

An object of class "cv.plsRmultiModel" with repeated fold predictions, observed fold responses, optional coefficient vectors and fold indices, and the reference full-data "plsRmultiModel" fit used for aggregated summary metrics.

Arguments

object: For the default method, a numeric multivariate response matrix or data frame with at least two columns. For the formula method, a formula of the form cbind(y1, y2, ...) ~ ..
...: Not used. Extra arguments are rejected in this experimental release.
dataX: Numeric predictor matrix or data frame.
nt: Number of components to extract in each fold fit.
limQ2set: Threshold used by cvtable for the aggregated Q2 criterion.
modele: Only "pls" is supported.
family: Not supported in this experimental release.
K: Number of groups for each partition.
NK: Number of repeated partitions.
grouplist: Optional user-supplied partitions.
random: Should the folds be generated randomly?
scaleX: Should predictors be scaled?
scaleY: Should responses be scaled? Defaults to TRUE.
keepcoeffs: Should standardized coefficient vectors be stored for each fold fit?
keepfolds: Should training indices be stored for each fold fit?
keepdataY: Kept for interface compatibility. Observed fold responses are stored so that summaries can be computed.
keepMclassed: Not supported in this experimental release.
EstimXNA: Not supported in this experimental release.
pvals.expli: Not supported in this experimental release.
alpha.pvals.expli: Not supported in this experimental release.
MClassed: Not supported in this experimental release.
tol_Xi: Tolerance used for degeneracy checks during component extraction.
weights: Not supported in this experimental release.
sparse: Not supported in this experimental release.
sparseStop: Not supported in this experimental release.
naive: Not supported in this experimental release.
verbose: Should informational messages be displayed?
data: An optional data frame for the formula method.
subset: An optional subset for the formula method.
contrasts: Optional contrasts for the formula method.

Details

Only the linear multivariate-response PLS2 mode is supported here. Missing values, weights, sparse extraction options, classification diagnostics, and GLM families remain out of scope for this experimental API.

Examples

Run this code

set.seed(123)
X <- matrix(rnorm(60 * 4), ncol = 4)
Y <- cbind(
  y1 = X[, 1] - 0.5 * X[, 2] + rnorm(60, sd = 0.1),
  y2 = 0.3 * X[, 2] + X[, 3] + rnorm(60, sd = 0.1)
)

cv_fit <- cv.plsRmulti(Y, X, nt = 2, K = 3, NK = 1, verbose = FALSE)
summary(cv_fit, verbose = FALSE)