get_rca.R

In jorgenfrost/compost: Compute Complexity Metrics

#' Creates an incidence matrix from a region-unit intensity matrix using RCA
#'
#' Takes a matrix with regions in rows and units in columns, where elements are
#' intensity. This could be countries as regions, products as units and export
#' value as intensity. Calculates the revealed comparative advantage for each
#' region-unit pair. Returns the incidence matrix.
#'
#' @param mat Matrix that contains regions on rows, units on columns and
#' intensity value in elements.
#' @param binary logical TRUE returns a binarized RCA matrix, where
#' values of at least 1 = 1, values below 1 = 0.
#' @return Matrix that contains the revealed comparative
#' of region (rows) in unit (columns).
#'
#' @export

get_rca <- function (mat, binary = TRUE) {
  # input: mat is a region x unit matrix where elements are intensity values
  # output: region x unit matrix where elements are rca values

  if (!is.matrix(mat)){
    stop("Input `mat` is not a matrix. Stopping.")
  }

  if (sum(is.na(mat)) != 0) {
    stop("Input `mat` contains NA values. Stopping.")
  }

    if (is.null(colnames(mat)) | is.null(rownames(mat))) {
        stop("`mat` seems to lack either colnames or rownames. Stopping.")
    }

  # numerator
  share_unit_region <- mat / rowSums(mat)

  # denominator
  share_unit_global <- colSums(mat) / sum(mat)

  incidence_mat <- t(t(share_unit_region) / share_unit_global)

  # NAs can be introduced in there are years where certain products are not exported at all (divide-by-zero problem)
  incidence_mat[is.na(incidence_mat)] <- 0

  if (binary == TRUE) {
    incidence_mat <- ifelse(incidence_mat >= 1, 1, 0)
  }

  return(incidence_mat)

}