R/utils.R

In adaptest: Data-Adaptive Statistics for High-Dimensional Multiple Testing

#' Print method for data_adapt objects
#'
#' Customized informative print method for examining data-adaptive statistics
#'
#' @param x (data_adapt) - object of class \code{data_adapt} as returned by
#' \code{adaptest}
#' @param ... additional arguments passed to \code{print} as necessary
#'
#' @return strings into stdout; containing information of the fitted model
#'
#' @export
#'
print.data_adapt <- function(x, ...) {
  print("The top covariates are")
  print(get_composition(x, type = "big")[[1]])
  print("The ATE estiamtes are")
  print(x$DE)
  print("The raw p-values are")
  print(x$p_value)
  print("The adjusted p-values are")
  print(x$q_value)
  print("The top mean CV-rank are (the smaller the better)")
  print(x$mean_rank_top)
  print(paste(
    "The percentage of appearing in top", length(x$top_colname),
    "are (the larger the better)"
  ))
  print(x$prob_in_top * 100)
  print("The covariates still significant are")
  print(x$significant_q)
  print("Their compositions are")
  print(get_composition(x, type = "small")[[1]])
}

################################################################################

#' Plot method for data_adapt objects
#'
#' Customized plotting method for easily examining data-adaptive statistics
#'
#' @param x (data_adapt) - object of class \code{data_adapt} as returned by
#' \code{adaptest}
#' @param plot_type character vector specifying which of the two types of plots
#'  to generate: "biomarker" for a plot sorted average CV-rank, or "adapt_param"
#'  for a plot sorted by q-values with labels corresponding to indices
#' @param ... additional arguments passed to \code{plot} as necessary
#'
#' @return plot of model statistics
#'
#' @importFrom graphics abline plot
#' @importFrom calibrate textxy
#'
#' @export
#
plot.data_adapt <- function(x, ..., plot_type = c("biomarker", "adapt_param")) {
  top_index <- x$top_index
  DE <- x$DE
  p_value <- x$p_value
  q_value <- x$q_value
  significant_q <- x$significant_q

  mean_rank_top <- x$mean_rank_top
  prob_in_top <- x$prob_in_top

  n_top.want <- length(top_index)

  if ("biomarker" %in% plot_type) {
    # Plot sorted average CV-rank
    plot(
      mean_rank_top,
      ylab = "Mean CV-rank", pch = 20,
      main = "Mean CV-rank of selected covariates \n (smaller is better)"
    )
    calibrate::textxy(
      (seq_len(n_top.want)) - 0.3, mean_rank_top + 0.5, top_index,
      offset = 0.6
    )
    abline(a = 0, b = 1, lty = 3)
  }

  if ("adapt_param" %in% plot_type) {
    # plot sorted q-values, labeled with index
    temp.top_index <- c(seq_len(n_top.want))[order(q_value)]
    plot(
      sort(q_value),
      pch = 20,
      ylab = "q-value",
      main = "q-value of selected covariates \n (smaller is better)"
    )
    calibrate::textxy(
      (seq_len(n_top.want))[seq_len(x$n_top)] - 0.3, sort(q_value),
      temp.top_index,
      offset = 1
    )
    abline(h = 0.05, lty = 2)
  }
}

################################################################################

#' Summary tables for data_adapt objects
#'
#' @param object (data_adapt) object as returned by \code{adaptest}
#' @param ... not implemented
#' @param type (character) - `adapt_param` or `biomarker`.
#'  `adapt_param` mode summarizes the data-adaptive target
#'  parameter. `biomarker` mode summarizes chracteristics of the biomarkers from
#'  the original data
#'
#' @return (data.frame) of the summary statistics
#' @return \code{type = 'adapt_param'} wtih columns: 'data-adaptive parameters',
#' 'Differential expression', 'p-values', 'q-values'
#' @return \code{type = 'biomarker'} wtih columns: 'biomakers', 'mean rank', '%
#' appear in top'
#'
#' @export
#' @method summary data_adapt
#'
summary.data_adapt <- function(object, type = "adapt_param", ...) {
  data_adapt_param <- seq_len(length(object$DE)) # 1:length(object$DE)
  DE <- object$DE
  p_value <- object$p_value
  q_value <- object$q_value

  top_biomarker <- object$top_index
  mean_rank_top <- object$mean_rank_top
  prob_in_top <- object$prob_in_top * 100

  if (type == "adapt_param") {
    table_out <- data.frame(data_adapt_param, DE, p_value, q_value)
    colnames(table_out) <- c(
      "data-adaptive parameters",
      "Differential expression", "p-values", "q-values"
    )
    print(table_out)
  }

  if (type == "biomarker") {
    len <- length(top_biomarker)
    table_out <- data.frame(top_biomarker, mean_rank_top, prob_in_top)
    colnames(table_out) <- c(
      "biomakers", "mean rank",
      paste("% appear in top ", len)
    )
    print(table_out)
  }
}