R/socialNetworkMeasures.r

In alex-naud/veritas-r: Toolkit for working with veritas data

#' Calculate social network measures
#'
#' This function calculate network measures from social networks 
#' generated from Veritas data.
#'
#' Calculate the following measures:
#'  - *people_degree:* Number of individually reported people
#'  - *group_degree:* Number of people generated from groups
#'  - *degree:* Number of people
#'  - *clustering:* Participant clustering coefficient. Equivalent of the
#'    network density without relations between the participants and the alters.
#'  - *simmelian:* Simmelian brokerage.
#'    See [Latora et al., 2013](https://doi.org/10.1007/s10955-013-0722-z)
#'
#' @param social_networks networks generated with [createSocialNetworks()]
#'
#' @return Data frame of network measures by participant
#'
#' @export
socialNetworkMeasures <- function(social_networks){

    # Create a function to calculate measure on a single network
    calculateMeasures <- function(g){

        # Degree
        people_degree <- sum(igraph::V(g)$status %in% "alter")
        group_degree <- sum(igraph::V(g)$status %in% "group")
        all_degree <- people_degree + group_degree

        # Clustering
        cc <- igraph::induced.subgraph(g, igraph::V(g)$status != "ego") %>%
                    igraph::graph.density()

        # Simmelian brokerage
        sb <- simmelianBrokerage(g)

        # Return
        c(people_degree = people_degree, group_degree = group_degree,
          degree = all_degree, clustering = cc, simmelian = sb)
    }

    # Run on all social networks
    out <- do.call(rbind, lapply(social_networks, calculateMeasures))

    # transform to data.frame and add pid as column
    out <- data.frame(out)
    out <- cbind(data.frame(pid = row.names(out)), out)
    row.names(out) <- NULL

    # Return
    return(out)
}

#' Calculate Simmelian Brokerage
#'
#' Adapted from a code by Christopher G. Watson.
#'  Retrieve from https://pastebin.com/XqkEYtJS
#'
#' @param g igraph network object generated with createSocialNetwork.r
#'
#' @return Simmelian Brokerage measure
#'
#' @noRd
simmelianBrokerage <- function(g){

  # Return NA if there is only less than two alter
  if(igraph::vcount(g) < 2) {
    return(NA)

    # Calculate local eff
  } else{

    # Remove ego
    g_sub <- igraph::induced.subgraph(g, igraph::V(g)$status != "ego")

    # Ego degree
    degree <- igraph::vcount(g_sub)

    # Calculate shortest path
    paths <- igraph::shortest.paths(g_sub, weights = NA)
    paths <- paths[upper.tri(paths)]

    # Calculate local efficiency
    local_eff <- 2 / degree / (degree - 1) * sum(1 / paths[paths != 0])

    # Calculate simmelian
    return(degree - (degree - 1) * local_eff)

  }
}