R/projections.R

In pachamaltese/economiccomplexity: Computational Methods for Economic Complexity

#' Projections of a Country-Product Network
#'
#' @description \code{projections()} computes two graphs that are particularly
#' useful to visualize product-product and country-country similarity.
#'
#' @details The current implementation follows
#' \insertCite{atlas2014}{economiccomplexity} to create simplified graphs
#' that correspond to a simplification of the proximity matrices. The result is
#' obtained by iterating and reducing links until the desired average number of
#' links per node is obtained, or a spaning tree after the strongest links is
#' returned when is not possible to return the desired network.
#'
#' @return A list of two graphs.
#'
#' @param proximity_country (Type: matrix) the output from
#' \code{proximity()}) or an equivalent arrangement.
#' @param proximity_product (Type: matrix) the output from
#' \code{proximity()}) or an equivalent arrangement.
#' @param avg_links average number of connections for the projections.
#' By default this is set to \code{5}.
#' @param tolerance tolerance for proximity variation on each iteration until
#' obtaining the desired average number of connections.
#' By default this is set to \code{0.05}.
#' @param compute (Type: character) the proximity to compute. By default this is
#' \code{"both"} (both projections) but it can also be \code{"country"}
#' or \code{"product"}.
#'
#' @examples
#' net <- projections(
#'   economiccomplexity_output$proximity$proximity_country,
#'   economiccomplexity_output$proximity$proximity_product,
#'   avg_links = 10,
#'   tolerance = 0.1
#' )
#'
#' # partial view of projections
#' n <- seq_len(5)
#' igraph::E(net$network_country)[n]
#' igraph::E(net$network_product)[n]
#'
#' @references
#' For more information see:
#'
#' \insertRef{atlas2014}{economiccomplexity}
#'
#' and the references therein.
#'
#' @keywords functions
#'
#' @export

projections <- function(proximity_country, proximity_product,
                        avg_links = 5, tolerance = 0.05, compute = "both") {
  # sanity checks ----
  if (!(any(class(proximity_country) %in% "matrix") == TRUE) |
    !(any(class(proximity_product) %in% "matrix") == TRUE)) {
    stop("'proximity_country' and 'proximity_product' must be matrix")
  }

  if (!is.numeric(avg_links)) {
    stop("'avg_links' must be numeric")
  }

  if (!any(compute %in% c("both", "country", "product"))) {
    stop("'compute' must be 'both', 'country' or 'product'")
  }

  trim_network <- function(proximity_mat, proximity_avg) {
    # this -1 is because the book by Hausmann mentions "maximum spanning tree"
    proximity_mat <- (-1) * proximity_mat

    g <- graph_from_adjacency_matrix(proximity_mat, weighted = TRUE,
      mode = "undirected", diag = FALSE)

    g_mst <- mst(g, algorithm = "prim")

    threshold <- 0
    avg_links_n <- FALSE

    while (avg_links_n == FALSE) {
      if (threshold < 1) {
        message(sprintf("%s threshold...", threshold))

        g_not_in_mst <- delete.edges(g, which(abs(E(g)$weight) <= threshold))
        g_not_in_mst <- graph.difference(g_not_in_mst, g_mst)

        g <- graph.union(g_mst, g_not_in_mst)
        E(g)$weight <- pmin(E(g)$weight_1, E(g)$weight_2, na.rm = TRUE)
        g <- remove.edge.attribute(g, "weight_1")
        g <- remove.edge.attribute(g, "weight_2")

        avg_links_n <- ifelse(mean(degree(g)) <= avg_links, TRUE, FALSE)
        threshold <- threshold + tolerance

        if (avg_links_n == TRUE) {
          message(sprintf("%s threshold achieves the avg number of connections",
            threshold))
          E(g)$weight <- (-1) * E(g)$weight
          return(g)
        }
      } else {
        warning(paste("no threshold achieves the avg number of connections",
                      "returning maximum spanning tree"))
        avg_links_n <- TRUE
        E(g_mst)$weight <- (-1) * E(g_mst)$weight
        return(g_mst)
      }
    }
  }

  if (any(compute == "country" || compute == "both") == TRUE) {
    message("computing country projection...")
    message(rep("-", 50))
    xg <- trim_network(proximity_country, avg_links)
  } else {
    xg <- NULL
  }

  if (any(compute == "product" || compute == "both") == TRUE) {
    message("computing product projection...")
    message(rep("-", 50))
    yg <- trim_network(proximity_product, avg_links)
  } else {
    yg <- NULL
  }

  return(
    list(
      network_country = xg,
      network_product = yg
    )
  )
}