R/alr.R

In OxWearables/epicoda: Supports epidemiological analyses using compositional exposure variables

#' Alr transformation
#'
#' This peforms additive log-ratio transformation.
#'
#' @param data Compositional columns of dataframe
#' @param comp_labels The labels of the compositional columns.
#' @param comparison_part Name of part that all other parts will be compared to.
#' @return \code{data} with alr-transformed compositional columns.
#' @examples alr_trans(data = simdata[,c("vigorous", "moderate", "light", "sedentary", "sleep")],
#' comp_labels = c("vigorous", "moderate", "light", "sedentary", "sleep"),
#' comparison_part = "sleep")
#'
#' @export
alr_trans <- function(data, comp_labels, comparison_part = NULL) {
  if (is.null(comparison_part)){
    stop("comparison_part must be specified for alr transformation. It is the part which all other parts will be expressed as ratios relative to.")
  }
  if (!(comparison_part %in% comp_labels)){
    stop("comparison_part must appear in comp_labels")
  }
  comp_labels_without_cc <- comp_labels[comp_labels != comparison_part]
  y <- data[, comp_labels_without_cc]
  for (label in comp_labels_without_cc){
    y[, label]<- log(y[, label]/data[, comparison_part])
  }
  colnames(y) <- paste0("alr_",colnames(y), comparison_part)
  return(y)
}