R/best_duplicate.R

In RemPsyc/rempsyc: Convenience Functions for Psychology

#' @title Choose the best duplicate
#'
#' @description Chooses the best duplicate, based on the
#' duplicate with the smallest number of missing values. In case of
#' ties, it picks the first duplicate, as it is the one most likely
#' to be valid and authentic, given practice effects.
#'
#' @details For the *easystats* equivalent, see:
#' [datawizard::data_duplicated()].
#' @param data The data frame.
#' @param id The ID variable for which to check for duplicates.
#' @param keep.rows Logical, whether to add a column at the beginning
#' of the data frame with the original row indices.
#' @keywords duplicates
#' @return A dataframe, containing only the "best" duplicates.
#' @export
#' @examples
#' df1 <- data.frame(
#'   id = c(1, 2, 3, 1, 3),
#'   item1 = c(NA, 1, 1, 2, 3),
#'   item2 = c(NA, 1, 1, 2, 3),
#'   item3 = c(NA, 1, 1, 2, 3)
#' )
#'
#' best_duplicate(df1, id = "id", keep.rows = TRUE)
#'
#' @importFrom dplyr mutate group_by group_by slice_min distinct %>% ungroup

best_duplicate <- function(data, id, keep.rows = FALSE) {
  check_col_names(data, id)

  og.names <- names(data)
  dups <- data %>%
    extract_duplicates(id)
  dups.n <- sum(duplicated(dups[[id]]))

  good.dups <- dups %>%
    group_by(.data[[id]]) %>%
    slice_min(.data$count_na) %>%
    distinct(.data[[id]], .keep_all = TRUE) %>%
    select(-all_of("count_na"))

  if (isTRUE(keep.rows)) {
    Row <- seq_len(nrow(data))
    data <- cbind(Row, data)
  } else {
    good.dups <- good.dups %>%
      select(-all_of("Row"))
  }

  good.data <- distinct(data, .data[[id]], .keep_all = TRUE)
  match.index <- good.data[[id]] %in% good.dups[[id]]
  good.data[match.index, ] <- good.dups

  dup.msg <- " duplicates removed)"
  dup.msg <- c(dup.msg, ifelse(dups.n != 69, "", " 69... nice"))
  message("(", dups.n, dup.msg, sep = "")
  good.data
}