R/format_bootstrap_list.R

In Tenenhaus/RGCCA: Regularized and Sparse Generalized Canonical Correlation Analysis for Multiblock Data

#' Format bootstrap results
#'
#' Raw bootstrap results are stored in a list of n_boot elements where each
#' element of this list is a list of J matrices. Each of these matrices has
#' ncomp[j] columns where ncomp[j] is the number of components for block j.
#' The aim of this function is to reorganize this list into a data.frame.
#' @param W raw bootstrap results
#' @param rgcca_res A fitted rgcca object.
#' @return A data.frame containing the results.
#' @noRd
format_bootstrap_list <- function(W, rgcca_res) {
  # Repeat values for variables, blocks and components
  grid <- do.call(rbind, lapply(seq_along(rgcca_res$a), function(j) {
    a <- rgcca_res$a[[j]]
    expand.grid(list(
      var = rownames(a),
      comp = seq_len(NCOL(a)),
      block = names(rgcca_res$a)[j]
    ))
  }))

  # Repeat values by adding the type and the number of the bootstrap sample
  df <- expand.grid(
    var = grid[, 1], type = c("weights", "loadings"), boot = seq_along(W)
  )
  df <- cbind(df, lapply(grid[-1], function(x) {
    expand.grid(
      x, type = c("weights", "loadings"), boot = seq_along(W)
    )[, 1]
  }))

  # Unlist the values into a new column of the grid
  df$value <- unlist(W, use.names = FALSE)

  return(df)
}