Transfer data from one domain/block to another via a latent space

Convert between data representations in a multiblock or cross-decomposition model by projecting the input new_data from the from domain/block onto a latent space and then reconstructing it in the to domain/block.

Usage

transfer(x, new_data, from, to, opts = list(), ...)

Arguments

x

The model fit, typically an object that implements a transfer method and ideally a block_names method.

new_data

The data to transfer, typically matching the dimension of the from domain.

from

Character string or index identifying the source domain/block. Must be present in block_names(x) if that method exists.

to

Character string or index identifying the target domain/block. Must be present in block_names(x) if that method exists.

opts

A list of optional arguments controlling the transfer process:

cols

Optional numeric vector specifying column indices of the target domain to reconstruct. If NULL (default), reconstructs all columns.

comps

Optional numeric vector specifying which latent components to use for the projection/reconstruction. If NULL (default), uses all components.

ls_rr

Logical; if TRUE, use a ridge-regularized LS approach for the initial projection from the from domain. Default FALSE.

lambda

Numeric ridge penalty (if ls_rr=TRUE). Default 1e-6.

...

Additional arguments passed to specific methods (discouraged, prefer opts).

Value

A matrix or data frame representing the transferred data in the to domain/block (or a subset of columns/components if specified in opts).