Reconstruct Data from an MFA Fit

MFA loadings are expressed in the block-weighted column metric (`components(x) = A `scores weights. These methods back-project with the metric-orthonormal axes `ov` instead, giving the rank-`length(comp)` generalized SVD approximation of the preprocessed data, and then undo the preprocessing.

Usage

# S3 method for class 'mfa'
reconstruct(
  x,
  comp = 1:multivarious::ncomp(x),
  rowind = 1:nrow(multivarious::scores(x)),
  colind = 1:nrow(multivarious::components(x)),
  ...
)

# S3 method for class 'mfa'
reconstruct_new(
  x,
  new_data,
  comp = 1:multivarious::ncomp(x),
  colind = 1:nrow(multivarious::components(x)),
  ...
)

Arguments

x

A fitted `mfa` object.

comp

Integer vector of components to use.

rowind

Training rows to reconstruct.

colind

Concatenated variables (columns) to reconstruct.

...

Unused.

new_data

Matrix of new observations: all concatenated variables, or only the variables in `colind`.

Value

A numeric matrix on the original data scale.