Partial Least Squares (PLS)

Fit a Partial Least Squares (PLS) model for regression (R) or discriminant analysis (DA), with automated cross-validated component selection.

Usage

pls(
  X,
  Y,
  center = TRUE,
  scale = "UV",
  cv = list(method = "k-fold_stratified", k = 7, split = 2/3),
  maxPCo = 5,
  plotting = TRUE
)

Arguments

X

A numeric matrix or data frame. Each row represents an observation, and each column a metabolic variable.

Y

Response vector or matrix. Must match the number of rows in X.

center

Logical. Should data be mean-centered? Default is TRUE.

scale

Character. Scaling method: "None", "UV" (unit variance), or "Pareto".

cv

Named list specifying cross-validation settings:

method

Cross-validation type: "k-fold", "k-fold_stratified", "MC", or "MC_balanced".

split

Fraction of observations used for training (used in Monte Carlo CV).

k

Number of folds or repetitions.

maxPCo

Integer. Maximum number of orthogonal components to test.

plotting

Logical. If TRUE, model summary (e.g. R2X, Q2, AUROC) is plotted. Default is TRUE.

Value

An object of class PLS_metabom8, an S4 class with scores, loadings, predictions, and validation statistics.

Details

Cross-validation is used to select the optimal number of predictive components based on Q2 or AUROC. The method supports both regression and classification with binary or multi-class responses. Model interpretability is often best with pairwise class comparisons.

References

Geladi, P. & Kowalski, B.R. (1986). Partial least squares regression: a tutorial. Analytica Chimica Acta, 185, 1–17.

Examples

data(covid)
X <- covid$X
an <- covid$an

model <- pls(X, Y = an$type)
#> Performing discriminant analysis.
#> Reducing k to 5 due to small group size (min n = 5).

plotscores(model, an = list(Class = an$type, Clinic = an$hospital, id = 1:nrow(an)), pc = c(1, 2))