Create a Frozen ComBat PipeOp for mlr3pipelines

Creates an mlr3pipelines PipeOp that wraps FrozenComBat for use in leakage-free ML pipelines. The batch correction parameters are learned during training and frozen for prediction.

Usage

create_frozen_combat_pipeop(
  batch_col = "batch",
  parametric = TRUE,
  mean_only = FALSE
)

Arguments

batch_col

Name of the batch column in the task (default: "batch")

parametric

Use parametric empirical Bayes (default: TRUE)

mean_only

Only correct means, not variances (default: FALSE)

Value

A PipeOp object

Examples

if (FALSE) { # \dontrun{
if (requireNamespace("mlr3learners", quietly = TRUE) &&
    requireNamespace("ranger", quietly = TRUE)) {
  # Create pipeline with frozen ComBat
  library(mlr3pipelines)

  po_combat <- create_frozen_combat_pipeop(batch_col = "batch")
  po_scale <- po("scale")
  po_learner <- po("learner", lrn("classif.ranger"))

  graph <- po_combat %>>% po_scale %>>% po_learner
}
} # }