Project DKGE data into component space — dkge_project

Functions for projecting subject-standardised betas or new blocks into DKGE component coordinates.

Convenience wrapper for projecting row-standardised betas onto DKGE components.

Usage

dkge_project_btil(fit, Btil)

dkge_project_block(
  fit,
  s,
  B_s,
  Omega_s = NULL,
  w_s = NULL,
  least_squares = TRUE
)

Arguments

fit: A `dkge` object.
Btil: Either a qxP matrix or a list of such matrices (e.g. `fit$Btil`).
s: Block index (subject) to project against.
B_s: Beta matrix for the new data block.
Omega_s: Optional weights (vector length P or PxP matrix) matching the columns of `B_s`.
w_s: Optional subject-level weight (defaults to 1 when omitted).
least_squares: Logical; pass to [multivarious::project_block()].

Value

List of Pxrank matrices; returns a single matrix when `Btil` is a matrix.

Projection scores (qxrank) restricted to block `s`.

Functions

dkge_project_btil(): Project subject-standardised betas into component space

Examples

toy <- dkge_sim_toy(
  factors = list(A = list(L = 2), B = list(L = 3)),
  active_terms = c("A", "B"), S = 3, P = 20, snr = 5
)
fit <- dkge_fit(toy$B_list, toy$X_list, toy$K, rank = 2)
A <- dkge_project_btil(fit, fit$Btil[[1]])
dim(A)
#> [1] 20  2