NIPALS PLS with missing values — nipals.fit • pls

A NIPALS implementation that tolerates NAs in both X and Y by ignoring them when updating scores and loadings. This is useful when the design matrix is incomplete but the number of components is relatively low.

Usage

nipals.fit(
  X,
  Y,
  ncomp,
  center = TRUE,
  stripped = FALSE,
  maxiter = 500,
  tol = 1e-06,
  ...
)

Arguments

X: numeric matrix (or coercible) of predictors. Missing values are allowed and handled internally.
Y: numeric matrix (or coercible) of responses. Missing values are also handled internally.
ncomp: number of PLS components to extract.
center: logical whether to center X and Y before fitting. Means ignore missing entries.
stripped: logical. If TRUE only the coefficients and the mean vectors are returned.
maxiter: maximum number of inner iterations to force convergence on each component.
tol: tolerance used to stop the inner loop when the direction vector changes very little.
...: currently ignored.

Value

A list with the same components as nipals.fit, but the computations never fail in the presence of missing entries.